goal.jsonl

{"task_id": "arxiv_001-g", "domain": "ARXIV", "autonomy_type": "ordered table", "oracle_output_cardinality": 4, "instruction": "I am organizing the research trajectory of dataset distillation and dataset condensation on arXiv from 2019 to 2021, aiming to identify machine learning papers that discuss using a small set of synthetic training data to replace large-scale training sets. From the 2019–2021 arXiv records, retain only entries whose primary category is Machine Learning; if a paper is only cross-listed into Machine Learning by another primary category, do not include it. Additionally, these papers must have a revision within the next calendar year after the initial submission, and the current record must still clearly indicate a subsequent publication venue. Output a table sorted by v1 submitted in ascending order, with columns: arXiv ID, term family, cue location, v1 submitted, first next-year version, date of that version, primary category, publication-trail source, current publication clue. If no qualifying entries exist, output NONE.", "start_url": "https://arxiv.org/search/advanced", "output_format": "Output a table sorted by v1 submitted in ascending order, with columns: arXiv ID, term family, cue location, v1 submitted, first next-year version, date of that version, primary category, publication-trail source, current publication clue. If no qualifying entries exist, output NONE.", "oracle_answer": "arXiv ID|term family|cue location|v1 submitted|first next-year version|date of that version|primary category|publication-trail source|current publication clue\n1910.02551|Distillation|Title|2019-10-06|v3|2020-05-05|cs.LG|Related DOI|10.1109/IJCNN52387.2021.9533769\n2011.00050|Distillation|Abstract-only|2020-10-30|v2|2021-03-02|cs.LG|Comments|Accepted to ICLR 2021\n2107.13034|Distillation|Title|2021-07-27|v3|2022-01-17|cs.LG|Comments|NeurIPS 2021\n2110.04181|Condensation|Title|2021-10-08|v2|2022-04-21|cs.LG|Journal reference|Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision 2023 (WACV)", "metadata": {"State-Gated Retrieval": ["Only include arXiv records where v1 submitted falls between 2019-01-01 and 2021-12-31.", "The title or abstract must explicitly use terms from the dataset distillation or dataset condensation family, and the semantics must refer to replacing large-scale training sets with a small set of synthetic training data.", "The current primary category must be cs.LG; if a paper is only cross-listed into Machine Learning by another primary category, do not include it.", "The paper must have a revision within the next calendar year after the initial submission, and the current record must still clearly indicate a subsequent publication venue."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["Root searches yield candidate arXiv IDs", "Each candidate abstract page yields primary category, v1 date, first next-year version, cue location, and publication-trail source", "Answer rows are produced only after deduplication and detail-page validation"], "control_dependency": ["Records must be filtered by original submission year, with per-record validation that a version in the next calendar year exists", "Cross-listed-only hits must be excluded, and the current primary category must be validated on abstract pages", "Searches must cover both Title and Abstract because valid distillation hits can appear only in the abstract", "Topic matches without a current publication trail still fail until the Comments, Journal reference, or Related DOI field is checked on the abstract page"], "freeze": {"historical_window": "v1 year in {2019, 2020, 2021}"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Only include arXiv records where v1 submitted falls between 2019-01-01 and 2021-12-31.", "The title or abstract must explicitly use terms from the dataset distillation or dataset condensation family, and the semantics must refer to replacing large-scale training sets with a small set of synthetic training data.", "The current primary category must be cs.LG; if a paper is only cross-listed into Machine Learning by another primary category, do not include it.", "The paper must have a revision within the next calendar year after the initial submission, and the current record must still clearly indicate a subsequent publication venue."], "exclusion_conditions": ["Exclude records whose current primary category is not cs.LG and that are only cross-listed to cs.LG.", "Exclude records where the title or abstract contains distillation/condensation terms but does not correspond to the dataset distillation or dataset condensation research line.", "Exclude records that have no version revision in the next calendar year, or whose current record contains no clearly visible publication clue."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "required literal first row shown in the oracle", "schema": ["arXiv ID", "term family", "cue location", "v1 submitted", "first next-year version", "date of that version", "primary category", "publication-trail source", "current publication clue"], "dedup_key": "bare arXiv ID without version suffix", "date_format": "YYYY-MM-DD", "term_family_extraction": {"allowed_families": ["Distillation", "Condensation"], "tie_break": "if both families appear, prefer the title hit; if both appear only in the abstract, keep the first abstract occurrence"}, "cue_location": {"Title": "qualifying cue appears only in the title", "Abstract-only": "qualifying cue appears only in the abstract", "Title + Abstract": "qualifying cue appears in both title and abstract, or title carries only a generic cue while the abstract carries the specific family name"}, "publication_trail_source": ["Comments", "Journal reference", "Related DOI"], "current_publication_clue": "keep the shortest phrase that still preserves venue or publication-status information; when sourced from Related DOI, drop the DOI prefix and keep only the informative clue", "sorting_or_selection": {"primary": "v1 submitted ascending", "secondary": "arXiv ID ascending"}, "stop_condition": ["all four corrected result pages are exhausted", "all unique candidate IDs are opened to abstract-page depth", "for every surviving candidate, the abstract page, current browse context, and publication-trail fields are checked before emitting the final ordered table"]}}}
{"task_id": "arxiv_002-g", "domain": "ARXIV", "autonomy_type": "ordered table", "oracle_output_cardinality": 10, "instruction": "I am compiling a reading list of 2022 research on reasoning prompting for large language models. From arXiv papers whose first version (v1) was submitted in 2022, I want to select those whose title or abstract explicitly focuses on chain-of-thought or self-consistency reasoning and whose primary category is Computation and Language. Candidate papers must also have a version update in 2023, and the current arXiv record's Comments field must directly indicate a subsequent publication venue. Output a table sorted by the date of each paper's earliest 2023 version in ascending order, with columns: arXiv ID, cue family, cue location, v1 submitted, first 2023 version, date of that version, primary category, publication clue. If no qualifying papers exist, output NONE.", "start_url": "https://arxiv.org/search/advanced", "output_format": "Output a table sorted by the date of each paper's earliest 2023 version in ascending order, with columns: arXiv ID, cue family, cue location, v1 submitted, first 2023 version, date of that version, primary category, publication clue. If no qualifying papers exist, output NONE.", "oracle_answer": "arXiv ID|cue family|cue location|v1 submitted|first 2023 version|date of that version|primary category|publication clue\n2210.01240|Chain-of-thought|Title + Abstract|2022-10-03|v2|2023-01-25|cs.CL|Published as a conference paper at ICLR 2023\n2205.11916|Chain-of-thought|Abstract-only|2022-05-24|v4|2023-01-29|cs.CL|Accepted to NeurIPS2022\n2203.11171|Self-consistency; Chain-of-thought|Title + Abstract|2022-03-21|v4|2023-03-07|cs.CL|Published at ICLR 2023; camera ready version at ICLR 2023\n2210.03629|Chain-of-thought|Abstract-only|2022-10-06|v3|2023-03-10|cs.CL|v3 is the ICLR camera ready version\n2210.03350|Chain-of-thought|Abstract-only|2022-10-07|v2|2023-05-23|cs.CL|To appear at Findings of EMNLP 2023\n2212.10001|Chain-of-thought|Title + Abstract|2022-12-20|v2|2023-06-01|cs.CL|ACL-23 Camera Ready\n2212.08061|Chain-of-thought|Abstract-only|2022-12-15|v2|2023-06-04|cs.CL|ACL 2023 Main Conference\n2212.10071|Chain-of-thought|Abstract-only|2022-12-20|v2|2023-06-13|cs.CL|ACL 2023 camera-ready\n2212.10509|Chain-of-thought|Title + Abstract|2022-12-20|v2|2023-06-23|cs.CL|ACL'23 Camera Ready\n2211.12588|Chain-of-thought; Self-consistency|Abstract-only|2022-11-22|v4|2023-10-23|cs.CL|Published at TMLR 2023", "metadata": {"State-Gated Retrieval": ["Only retain arXiv papers whose v1 was submitted in 2022.", "The title or abstract must explicitly contain cues related to chain-of-thought or self-consistency.", "The current primary category must be cs.CL; do not include records that are only cross-listed from other primary categories.", "The paper must have a version update in 2023, and the current Comments field must clearly indicate a subsequent publication venue."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["root searches yield candidate arXiv IDs", "each candidate abstract page yields primary category, v1 date, first 2023 version, comments/publication clue", "answer rows are produced only after deduplication and detail-page validation"], "control_dependency": ["searches must cover both Title and Abstract fields because valid rows can appear only in the abstract", "cross-listed hits whose primary category is not cs.CL must be excluded, and the primary category must be validated on the abstract page", "topic-matched hits must be validated against a current Comments publication clue before they can be accepted", "2023 candidates must be identified by original submission year and full version-history checks, rather than by latest-date proxies"], "freeze": {"historical_window": "v1 submitted in 2022; version-history event in 2023"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Only retain arXiv papers whose v1 was submitted in 2022.", "The title or abstract must explicitly contain cues related to chain-of-thought or self-consistency.", "The current primary category must be cs.CL; do not include records that are only cross-listed from other primary categories.", "The paper must have a version update in 2023, and the current Comments field must clearly indicate a subsequent publication venue."], "exclusion_conditions": ["Exclude records whose current primary category is not cs.CL, even if they appear in relevant search results.", "Exclude records that contain only vague abbreviations or weak related hints, but whose title and abstract lack clear CoT/self-consistency semantics.", "Exclude records that have no 2023 version update, or whose current record lacks a direct publication clue."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "required literal first row shown in the oracle", "schema": ["arXiv ID", "cue family", "cue location", "v1 submitted", "first 2023 version", "date of that version", "primary category", "publication clue"], "dedup_key": "bare arXiv ID without version suffix", "date_format": "YYYY-MM-DD", "cue_family": ["Chain-of-thought", "Self-consistency"], "cue_location": {"Title": "qualifying cue appears only in the title", "Abstract-only": "qualifying cue appears only in the abstract", "Title + Abstract": "qualifying cue appears in both title and abstract, or title carries only a generic cue while the abstract carries the specific family name"}, "publication_clue": {"source": "Comments field only", "extraction": "keep the shortest phrase that still preserves the current publication-status clue"}, "sorting_or_selection": {"primary": "date of the earliest 2023 version ascending", "secondary": "arXiv ID ascending"}, "stop_condition": ["all corrected result pages are exhausted", "all unique candidate IDs are opened to abstract-page depth", "for every surviving candidate, the current browse context, Comments field, and full Submission history are checked before emitting the final ordered table"]}}}
{"task_id": "arxiv_003-g", "domain": "ARXIV", "autonomy_type": "ordered table", "oracle_output_cardinality": 5, "instruction": "I am compiling a reading list of early vision Transformer papers. From arXiv papers whose first version (v1) was submitted in Q4 2020, I want to select those whose title or abstract explicitly uses vision- or image-related Transformer terminology and whose primary category is Computer Vision and Pattern Recognition. Additionally, eligible papers must have had at least one updated version in 2021, and their current arXiv record must still directly indicate a subsequent publication venue. Output a table sorted in ascending order by the date of each paper's earliest 2021 version, with columns: arXiv ID, cue family, cue location, v1 submitted, first 2021 version, date of that version, primary category, publication-trail source, current publication clue. If no qualifying papers exist, output NONE.", "start_url": "https://arxiv.org/search/advanced", "output_format": "Output a table sorted in ascending order by the date of each paper's earliest 2021 version, with columns: arXiv ID, cue family, cue location, v1 submitted, first 2021 version, date of that version, primary category, publication-trail source, current publication clue. If no qualifying papers exist, output NONE.", "oracle_answer": "arXiv ID|cue family|cue location|v1 submitted|first 2021 version|date of that version|primary category|publication-trail source|current publication clue\n2101.01097|Vision Transformer|Abstract-only|2020-12-30|v2|2021-01-08|cs.CV|Journal reference|IEEE Int. Conf. on Image Processing (ICIP) 2021\n2012.12556|Visual Transformer; Vision Transformer|Title + Abstract|2020-12-23|v2|2021-01-15|cs.CV|Comments; Related DOI|Accepted by TPAMI 2022; 10.1109/TPAMI.2022.3152247\n2012.00364|Image Processing Transformer|Title + Abstract|2020-12-01|v3|2021-05-28|cs.CV|Comments|CVPR 2021\n2011.08019|Vision Transformer|Title + Abstract|2020-11-16|v2|2021-06-02|cs.CV|Comments|Accepted for Publication in IJCB2021\n2010.11929|Vision Transformer|Abstract-only|2020-10-22|v2|2021-06-03|cs.CV|Comments|ICLR camera-ready version", "metadata": {"State-Gated Retrieval": ["Only include arXiv papers whose v1 submission date falls in Q4 2020.", "The title or abstract must explicitly contain vision- or image-related Transformer terminology.", "The current primary category must be cs.CV.", "The paper must have had an updated version in 2021, and the current record must still directly indicate a subsequent publication venue."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["root searches yield candidate arXiv IDs", "candidate abstract pages yield v1 submitted date, first 2021 version, primary category, cue location, and publication-trail source", "answer rows depend on deduplication plus detail-page validation"], "control_dependency": ["searches must cover both Title and Abstract because valid positives can appear only in the abstract", "candidate eligibility must be determined by the original-submission date window rather than by arXiv-ID or announcement intuition", "candidate filtering must use full Submission-history validation rather than most-recent-submission filtering", "searches must use the broader semantic cue state rather than a ViT-only entry point, because ViT-only hits can be false positives", "publication-trail false positives still require Comments / Journal reference / Related DOI validation before final inclusion"], "freeze": {"historical_window": "v1 submitted from 2020-10-01 through 2020-12-31; version-history event in 2021"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Only include arXiv papers whose v1 submission date falls in Q4 2020.", "The title or abstract must explicitly contain vision- or image-related Transformer terminology.", "The current primary category must be cs.CV.", "The paper must have had an updated version in 2021, and the current record must still directly indicate a subsequent publication venue."], "exclusion_conditions": ["Exclude records whose current primary category is not cs.CV and are only cross-listed.", "Exclude records that contain only weak cues such as Transformer/ViT but lack explicit vision- or image-related terminology in the title and abstract.", "Exclude records that had no version update in 2021 or whose current record contains no direct publication clue."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "required literal first row shown in the oracle", "schema": ["arXiv ID", "cue family", "cue location", "v1 submitted", "first 2021 version", "date of that version", "primary category", "publication-trail source", "current publication clue"], "dedup_key": "bare arXiv ID without version suffix", "date_format": "YYYY-MM-DD", "cue_family_extraction": "record every qualifying family actually present and join multiple families with `; ` in the emitted field", "cue_location": {"Title": "qualifying cue appears only in the title", "Abstract-only": "qualifying cue appears only in the abstract", "Title + Abstract": "qualifying cue appears in both title and abstract, or title carries only a generic cue while the abstract carries the specific family name"}, "publication_trail_source": ["Comments", "Journal reference", "Related DOI"], "current_publication_clue": "keep the shortest phrase that still preserves venue or publication-status information; when sourced from Related DOI, drop the DOI prefix and keep only the informative clue", "sorting_or_selection": {"primary": "date of the earliest 2021 version ascending", "secondary": "arXiv ID ascending"}, "stop_condition": ["all corrected result pages are exhausted", "all unique candidate IDs are opened to abstract-page depth", "for every surviving candidate, the abstract page, current browse context, Comments field, Journal reference, Related DOI, and full Submission history are checked before emitting the final ordered table"]}}}
{"task_id": "arxiv_004-g", "domain": "ARXIV", "autonomy_type": "ordered table", "oracle_output_cardinality": 5, "instruction": "From arXiv papers whose first version was submitted in the first half of 2020, select those whose primary category is Machine Learning and whose title or abstract clearly focuses on certified robustness defenses. The research subject of each candidate paper must explicitly involve at least one of the following cues: randomized smoothing, de-randomized smoothing, transformation-specific smoothing, smoothed classifiers, certified defenses against backdoor/poisoning, or partition aggregation. Additionally, these papers must have a version update in 2021, and the current arXiv record must still show a formal publication or acceptance destination. Output a table sorted in ascending order by the date of each paper's earliest 2021 version, with columns: arXiv ID, defense family, threat/topic cue, cue location, v1 submitted, first 2021 version, date of that version, primary category, publication clue. If no qualifying items exist, output NONE.", "start_url": "https://arxiv.org/search/advanced", "output_format": "Output a table sorted in ascending order by the date of each paper's earliest 2021 version, with columns: arXiv ID, defense family, threat/topic cue, cue location, v1 submitted, first 2021 version, date of that version, primary category, publication clue. If no qualifying items exist, output NONE.", "oracle_answer": "arXiv ID|defense family|threat/topic cue|cue location|v1 submitted|first 2021 version|date of that version|primary category|publication clue\n2002.10733|De-randomized smoothing; Randomized smoothing|patch attacks|Title + Abstract|2020-02-25|v3|2021-01-08|cs.LG|NeurIPS 2020\n2006.04062|Smoothed classifiers; Randomized smoothing|certified L2 robustness|Title + Abstract|2020-06-07|v4|2021-01-08|cs.LG|NeurIPS 2020\n2003.08904|RAB; Randomized smoothing|backdoor attacks|Title + Abstract|2020-03-19|v4|2021-02-26|cs.LG|IEEE Symposium on Security and Privacy 2023\n2006.14768|Deep Partition Aggregation|general poisoning attacks; label-flipping poisoning attacks|Title + Abstract|2020-06-26|v2|2021-03-18|cs.LG|ICLR 2021\n2002.12398|Transformation-specific smoothing; Randomized smoothing|semantic transformations|Title + Abstract|2020-02-27|v4|2021-09-17|cs.LG|2021 ACM SIGSAC Conference on Computer and Communications Security (CCS '21)", "metadata": {"State-Gated Retrieval": ["Only include arXiv papers with v1 submitted between 2020-01-01 and 2020-06-30.", "The title or abstract must clearly focus on randomized smoothing, de-randomized smoothing, transformation-specific smoothing, smoothed classifiers, certified defenses against backdoor or poisoning attacks, or partition aggregation.", "The current primary category must be cs.LG.", "Must have a version update in 2021, and the current record must show a formal publication or acceptance destination."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["root searches yield candidate arXiv IDs", "candidate abstract pages yield v1 date, primary category, full Submission history, cue location, and current publication clue", "final rows depend on deduplication plus detail-page validation"], "control_dependency": ["candidate filtering must use the original-submission window together with full 2021 version-history validation, because valid rows can have later newest versions", "cross-listed hits whose primary category is not the target category must be excluded, with category validation on detail pages", "searches must cover the broader certified-defense family, including partition-aggregation and backdoor/poisoning variants, rather than smoothing-only entries", "topic+venue hits must be confirmed by detail-history validation that a 2021 version actually exists"], "freeze": {"historical_window": "v1 submitted from 2020-01-01 through 2020-06-30; version-history event in 2021"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Only include arXiv papers with v1 submitted between 2020-01-01 and 2020-06-30.", "The title or abstract must clearly focus on randomized smoothing, de-randomized smoothing, transformation-specific smoothing, smoothed classifiers, certified defenses against backdoor or poisoning attacks, or partition aggregation.", "The current primary category must be cs.LG.", "Must have a version update in 2021, and the current record must show a formal publication or acceptance destination."], "exclusion_conditions": ["Exclude records whose current primary category is not cs.LG, even if semantically related.", "Exclude records that are only broadly related to robustness, defense, or backdoor/poisoning but do not meet the required certified-defense semantics.", "Exclude records without a 2021 version update or without a current publication clue."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "required literal first row shown in the oracle", "schema": ["arXiv ID", "defense family", "threat/topic cue", "cue location", "v1 submitted", "first 2021 version", "date of that version", "primary category", "publication clue"], "dedup_key": "bare arXiv ID without version suffix", "date_format": "YYYY-MM-DD", "defense_family_extraction": "record every qualifying defense family actually present and join multiple families with `; ` in the emitted field", "threat_topic_cue": "keep the shortest qualifying threat or topic cue that justifies inclusion", "cue_location": {"Title": "qualifying cue appears only in the title", "Abstract-only": "qualifying cue appears only in the abstract", "Title + Abstract": "qualifying cue appears in both title and abstract, or title carries only a generic cue while the abstract carries the specific family name"}, "publication_clue": "keep the shortest phrase that still preserves venue or publication-status information", "sorting_or_selection": {"primary": "date of the earliest 2021 version ascending", "secondary": "arXiv ID ascending"}, "stop_condition": ["all corrected result pages are exhausted", "all unique candidate IDs are opened to abstract-page depth", "for every surviving candidate, the abstract page, current browse context, Comments field, and full Submission history are checked before emitting the final ordered table"]}}}
{"task_id": "census_001-g", "domain": "CENSUS_GOV", "autonomy_type": "ordered table", "oracle_output_cardinality": 20, "instruction": "Using the 2014–2018 American Community Survey 5-year estimates, identify North Carolina counties that simultaneously exhibit insufficient internet access, a high share of seniors living alone, limited transportation options, and low income. Candidate counties must meet all of the following conditions: at least 4,000 households with no internet access; a share of households with no internet access higher than the statewide share; a share of households with a householder aged 65 or older living alone higher than the statewide share; a share of households with no vehicle higher than the statewide share; and a median household income lower than the statewide median. All statewide comparisons must use the North Carolina statewide estimate from the same data release. List all qualifying counties in a single line ordered alphabetically by county name. Each county should be formatted as County|NoInternetHH|NoInternetRatePct|SeniorLivingAloneSharePct|NoVehicleRatePct|MedianIncome, with counties separated by commas. Percentage fields should retain three decimal places without a percent sign.", "start_url": "https://www.census.gov/data.html", "output_format": "List all qualifying counties in a single line ordered alphabetically by county name. Each county should be formatted as County|NoInternetHH|NoInternetRatePct|SeniorLivingAloneSharePct|NoVehicleRatePct|MedianIncome, with counties separated by commas. Percentage fields should retain three decimal places without a percent sign.", "oracle_answer": "Beaufort County|5988|30.986|16.083|7.664|43688,Bladen County|4194|30.026|13.681|10.417|32378,Cleveland County|11599|31.449|12.469|7.825|40393,Columbus County|7371|33.045|15.148|7.958|36398,Duplin County|7574|34.773|14.370|7.424|38850,Edgecombe County|7454|34.879|13.720|11.225|35516,Halifax County|7575|35.873|14.657|11.565|34251,Lenoir County|4923|21.292|15.614|11.699|38387,Nash County|8617|23.467|13.597|7.917|48362,Richmond County|5193|28.001|13.706|9.269|36091,Robeson County|19814|43.103|11.312|9.854|33679,Rockingham County|9938|26.714|14.166|7.164|42490,Rutherford County|7583|28.783|16.033|6.335|40347,Sampson County|8216|34.907|14.088|7.686|39068,Scotland County|4971|37.909|13.910|11.470|35617,Surry County|7595|26.203|13.762|6.779|41068,Vance County|5131|30.444|12.347|11.600|37282,Wayne County|9709|20.163|11.742|8.456|42192,Wilkes County|7046|24.849|14.773|5.992|40829,Wilson County|7652|23.817|13.322|9.540|42850", "metadata": {"State-Gated Retrieval": ["Use only the 2014-2018 American Community Survey 5-year estimates for North Carolina counties, and use the North Carolina statewide value from the same release as the comparison baseline.", "Retain only counties with at least 4,000 households with no internet access, and where the share of households with no internet access, the share of households with a householder aged 65 or older living alone, and the share of households with no vehicle are all higher than the statewide shares, while the median household income is lower than the statewide median.", "Output the final results in alphabetical order by county name, retaining the six required fields."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the 2014-2018 ACS 5-year product and North Carolina county geography yield the full county candidate set plus the statewide baseline row", "the relevant ACS tables yield no-internet households, senior-living-alone share, no-vehicle share, and median household income for each county and the state baseline", "final rows are produced only after the 5-year product/geography setup is fixed consistently and the full county filtering chain is executed from B28002 through the downstream filters"], "control_dependency": ["the 2018 ACS 5-year Detailed Tables product must be used instead of the 2018 ACS 1-year product", "county coverage must be established under one consistent product-and-geography setup rather than by piecemeal fixes within the current table", "all downstream county comparisons must use the same 5-year statewide baseline and the same all-counties geography"], "freeze": {"historical_window": "2014-2018 American Community Survey 5-year estimates for North Carolina counties, using the North Carolina statewide row from the same release as the comparison baseline"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Use only the 2014-2018 American Community Survey 5-year estimates for North Carolina counties, and use the North Carolina statewide value from the same release as the comparison baseline.", "Retain only counties with at least 4,000 households with no internet access, and where the share of households with no internet access, the share of households with a householder aged 65 or older living alone, and the share of households with no vehicle are all higher than the statewide shares, while the median household income is lower than the statewide median.", "Output the final results in alphabetical order by county name, retaining the six required fields."], "exclusion_conditions": ["Exclude results that incorrectly use the 2018 ACS 1-Year Estimates Detailed Tables or any data other than the 2014-2018 5-year release.", "Exclude results that do not use the North Carolina statewide row from the same release as the comparison baseline.", "Exclude any county that does not satisfy every threshold or statewide comparison condition."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["County", "NoInternetHH", "NoInternetRatePct", "SeniorLivingAloneSharePct", "NoVehicleRatePct", "MedianIncome"], "county": "use the county display name including the word `County` and excluding a trailing state suffix", "dedup_key": "county FIPS / GEOID", "percentage_fields": "all percentage fields are plain numbers with exactly three decimal places and no percent sign", "numeric_fields": "household counts and median income are emitted as plain integers without commas", "sorting_or_selection": "county name alphabetical ascending", "stop_condition": ["the North Carolina statewide threshold is fixed once from the designated table", "all 100 North Carolina counties are covered under the same table/product settings", "every surviving county is validated against all four required tables before emission"]}}}
{"task_id": "census_002-g", "domain": "CENSUS_GOV", "autonomy_type": "ordered table", "oracle_output_cardinality": 5, "instruction": "Using the 2015–2019 American Community Survey 5-year estimates, identify New Mexico counties that simultaneously exhibit large renter populations, notable language barriers, limited internet access, transportation disadvantages, low income, and rental stress. These counties are intended to assess which areas were unsuitable for replacing in-person tenant legal aid with online self-service before the pandemic. Candidate counties must meet all of the following conditions: at least 2,500 renter-occupied households; at least 500 limited-English-speaking households; a share of limited-English-speaking households higher than the statewide share; a share of households with no internet access higher than the statewide share; a share of households with no vehicle higher than the statewide share; a median household income lower than the statewide median; and rental stress higher than the statewide level. Rental stress higher than the statewide level must be demonstrated by at least one of the following: a share of renter households with a rent-to-income ratio of 35% or higher that exceeds the statewide share, or a share of renter households with more than one occupant per room that exceeds the statewide share. All comparisons of “higher than statewide” or “lower than statewide” use the New Mexico statewide value from the same data release. Output a single line with all qualifying counties sorted alphabetically by county name. Each county is written as County|RenterHH|LEPHH|LEPSharePct|NoInternetPct|NoVehiclePct|SevereRentBurdenPct|CrowdedRenterPct|MedianIncome|StressPath, where StressPath must be RENT, CROWD, or BOTH. Separate multiple counties with commas. Percentage fields retain three decimal places with no percent sign.", "start_url": "https://data.census.gov/advanced", "output_format": "Output a single line with all qualifying counties sorted alphabetically by county name. Each county is written as County|RenterHH|LEPHH|LEPSharePct|NoInternetPct|NoVehiclePct|SevereRentBurdenPct|CrowdedRenterPct|MedianIncome|StressPath, where StressPath must be RENT, CROWD, or BOTH. Separate multiple counties with commas. Percentage fields retain three decimal places with no percent sign.", "oracle_answer": "Doña Ana County|28729|7883|10.127|22.612|6.079|44.172|6.391|40973|BOTH,Luna County|3482|1109|12.455|34.030|8.165|42.700|1.407|29360|RENT,McKinley County|6090|1610|7.688|50.587|10.978|27.568|16.256|33834|CROWD,San Miguel County|3443|1200|10.337|38.470|7.684|47.376|2.440|30946|RENT,Taos County|2854|652|5.387|28.844|5.726|48.823|4.555|38329|RENT", "metadata": {"State-Gated Retrieval": ["Use only New Mexico county data from the 2015–2019 American Community Survey 5-year estimates, and use the New Mexico statewide value from the same release as the comparison baseline.", "Candidate counties must have at least 2,500 renter households, at least 500 limited-English-speaking households, and LEP share, no-internet share, and no-vehicle share all higher than the statewide values, while median household income is lower than the statewide value.", "LEPHH must be computed by summing all four limited-English language groups in C16002. Retain only counties where at least one of Severe Rent Burden or Crowded Renter exceeds the statewide value, and write StressPath according to the rule."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the 2015-2019 ACS 5-year county geography yields New Mexico county candidates and the statewide baseline row", "C16002 and the downstream housing/income tables yield LEPHH, LEP share, no-internet share, no-vehicle share, severe rent burden, crowded-renter share, and median income for counties and the state baseline", "final rows are produced only after the LEPHH numerator is computed from all four limited-English language groups and the downstream stress-path filters are applied"], "control_dependency": ["the LEPHH numerator must be computed from all four limited-English language groups in C16002 rather than from Spanish-only counts", "county eligibility and StressPath labels must be determined from the LEPHH measure and statewide LEP baseline defined under that full four-group calculation", "the RENT / CROWD / BOTH classification must be determined by evaluating the same county set against both downstream housing-stress branches"], "freeze": {"historical_window": "2015-2019 American Community Survey 5-year estimates for New Mexico counties, using the New Mexico statewide row from the same release as the comparison baseline"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Use only New Mexico county data from the 2015–2019 American Community Survey 5-year estimates, and use the New Mexico statewide value from the same release as the comparison baseline.", "Candidate counties must have at least 2,500 renter households, at least 500 limited-English-speaking households, and LEP share, no-internet share, and no-vehicle share all higher than the statewide values, while median household income is lower than the statewide value.", "LEPHH must be computed by summing all four limited-English language groups in C16002. Retain only counties where at least one of Severe Rent Burden or Crowded Renter exceeds the statewide value, and write StressPath according to the rule."], "exclusion_conditions": ["Exclude results that treat Spanish-only limited-English households as the entire LEPHH count.", "Exclude results that do not use the statewide baseline from the same release or do not recompute LEPHH/LEPSharePct/LEPShare_state.", "Exclude counties that do not satisfy the final stress-path condition of RENT, CROWD, or BOTH."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["County", "RenterHH", "LEPHH", "LEPSharePct", "NoInternetPct", "NoVehiclePct", "SevereRentBurdenPct", "CrowdedRenterPct", "MedianIncome", "StressPath"], "county": "use the county display name including the word `County` and excluding a trailing state suffix", "dedup_key": "county FIPS / GEOID", "LEPHH": "sum the four designated limited-English-proficient renter-household cells before emitting the integer total", "percentage_fields": "all percentage fields are plain numbers with exactly three decimal places and no percent sign", "numeric_fields": "household counts and median income are emitted as plain integers without commas", "StressPath": {"allowed_values": ["RENT", "CROWD", "BOTH"], "rule": "emit BOTH only when both the severe-rent-burden and crowded-renter signals survive; otherwise emit the single surviving label"}, "sorting_or_selection": "county name alphabetical ascending", "stop_condition": ["all candidate counties are validated under the same geography and product settings", "the renter scope and LEP aggregation are checked before the final line is emitted"]}}}
{"task_id": "census_003-g", "domain": "CENSUS_GOV", "autonomy_type": "ordered table", "oracle_output_cardinality": 4, "instruction": "Using the 2013–2017 American Community Survey 5-year estimates for California counties, identify counties where, for school meal material updates and annual parent information updates, relying solely on an English-only online portal is insufficient and paper materials with offline assistance must still be retained. Candidate counties must meet all of the following conditions: population aged 5–17 at least 75,000; among the population aged 5–17, the number of people who speak English \"not well\" or \"not at all\" is at least 2,000 and its share is higher than the statewide share; share of households without internet access higher than the statewide share; median household income lower than the statewide share; and at least two of three support-need signals higher than the statewide share. The three support-need signals are: share of households with no vehicle; share of minors living in households that received SSI, cash public assistance, or Food Stamps/SNAP in the past 12 months; and share of renter-occupied households with more than one occupant per room. All comparisons of \"higher than statewide\" or \"lower than statewide\" use the California statewide value from the same data release. Output all qualifying counties in a single line sorted alphabetically by county name. Each county is written as County|SchoolAgePop|SchoolAgeLEP|LEPSharePct|NoInternetPct|NoVehiclePct|ChildBenefitSharePct|CrowdedRenterPct|MedianIncome|SupportPath, where SupportPath uses only the three labels BENEFITS, CROWD, and MOBILITY joined by + in that fixed order. Multiple counties are separated by commas.", "start_url": "https://data.census.gov/advanced", "output_format": "Output all qualifying counties in a single line sorted alphabetically by county name. Each county is written as County|SchoolAgePop|SchoolAgeLEP|LEPSharePct|NoInternetPct|NoVehiclePct|ChildBenefitSharePct|CrowdedRenterPct|MedianIncome|SupportPath, where SupportPath uses only the three labels BENEFITS, CROWD, and MOBILITY joined by + in that fixed order. Multiple counties are separated by commas.", "oracle_answer": "Fresno County|200205|5127|2.561|20.588|8.686|46.388|13.967|48730|BENEFITS+CROWD+MOBILITY,Kern County|185802|4130|2.223|21.739|7.058|37.285|14.394|50826|BENEFITS+CROWD,Los Angeles County|1641669|37540|2.287|16.338|9.204|27.238|16.741|61015|BENEFITS+CROWD+MOBILITY,Tulare County|104542|2703|2.586|25.738|6.274|49.347|14.535|44871|BENEFITS+CROWD", "metadata": {"State-Gated Retrieval": ["Use only California county data from the 2013–2017 American Community Survey 5-year estimates, and use the California statewide value from the same release as the comparison baseline.", "Retain only counties with a population aged 5–17 of at least 75,000, a SchoolAgeLEP count of at least 2,000 with its share higher than the statewide share, a no-internet share higher than the statewide share, and a median household income lower than the statewide share.", "SchoolAgeLEP must be the sum of the eight not-well/not-at-all cells across the four major language groups in B16004. The final county must also have at least two of the three offline support signals (Mobility, Benefits, Crowd) higher than the statewide share."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the 2013-2017 ACS 5-year county geography yields California county candidates and the statewide baseline row", "B16004 and the downstream access/income/support tables yield SchoolAgeLEP, no-internet share, no-vehicle share, child-benefit share, crowded-renter share, and median income for counties and the statewide baseline", "final rows are produced only after the SchoolAgeLEP numerator is computed from all eight 1980-target language cells and the downstream support-path branches are applied"], "control_dependency": ["the SchoolAgeLEP numerator must be computed from the full eight-cell language set in B16004 rather than from only two Spanish LEP cells", "county eligibility and SupportPath labels must be determined from the LEP numerator and candidate set defined under that full eight-cell calculation", "the BENEFITS / CROWD / MOBILITY path labels must be determined by applying all three downstream support branches to the same county set"], "freeze": {"historical_window": "2013-2017 American Community Survey 5-year estimates for California counties, using the California statewide row from the same release as the comparison baseline"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Use only California county data from the 2013–2017 American Community Survey 5-year estimates, and use the California statewide value from the same release as the comparison baseline.", "Retain only counties with a population aged 5–17 of at least 75,000, a SchoolAgeLEP count of at least 2,000 with its share higher than the statewide share, a no-internet share higher than the statewide share, and a median household income lower than the statewide share.", "SchoolAgeLEP must be the sum of the eight not-well/not-at-all cells across the four major language groups in B16004. The final county must also have at least two of the three offline support signals (Mobility, Benefits, Crowd) higher than the statewide share."], "exclusion_conditions": ["Exclude results that use only the two Spanish LEP rows without merging all eight target cells.", "Exclude results that do not rerun the B28011, B19013, B08201, B09010, and B25014 branches.", "Exclude counties whose final SupportPath does not have at least two signals higher than the statewide share."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["County", "SchoolAgePop", "SchoolAgeLEP", "LEPSharePct", "NoInternetPct", "NoVehiclePct", "ChildBenefitSharePct", "CrowdedRenterPct", "MedianIncome", "SupportPath"], "county": "use the county display name including the word `County` and excluding a trailing state suffix", "dedup_key": "county FIPS / GEOID", "SchoolAgeLEP": "sum the eight designated school-age LEP cells before emitting the integer total", "percentage_fields": "all percentage fields are plain numbers with exactly three decimal places and no percent sign", "numeric_fields": "population counts and median income are emitted as plain integers without commas", "SupportPath": "use only BENEFITS, CROWD, and MOBILITY, and when multiple labels apply join them with `+` in the fixed order BENEFITS+CROWD+MOBILITY", "sorting_or_selection": "county name alphabetical ascending", "stop_condition": ["all candidate counties are validated under the same geography and product settings", "the school-age LEP aggregation and all three support-signal gates are checked before the final line is emitted"]}}}
{"task_id": "census_004-g", "domain": "CENSUS_GOV", "autonomy_type": "ordered table", "oracle_output_cardinality": 7, "instruction": "Using the 2012–2016 American Community Survey 5-year estimates, identify Pennsylvania counties with high elderly evacuation support needs that should conduct in-person outreach before winter storms. A candidate county must meet all of the following conditions: a population aged 65 and older of at least 15,000; a population aged 65 and older with ambulatory difficulty of at least 4,000, with its share of the 65+ population higher than the statewide share; a median household income below the statewide median; and at least two of four evacuation-support barriers above the statewide level. The four barriers are: share of households with a householder aged 65+ living alone, share of households with no vehicle, share of renter-occupied households with no vehicle, and mobile homes as a share of all housing units. All comparisons to the statewide level use the Pennsylvania statewide values from the same data release. Output all qualifying counties as a single line, sorted alphabetically by county name. Each county is formatted as County|SeniorPop|AmbulatoryDifficulty65Plus|AmbDiffPct|SeniorLivingAloneSharePct|NoVehiclePct|RenterNoVehiclePct|MobileHomePct|MedianIncome|BarrierPath, with counties separated by commas. For BarrierPath, use only ALONE, MOBILE_HOME, NO_VEHICLE, and RENTER_NO_VEHICLE, joined by + in that fixed order.", "start_url": "https://www.census.gov/data.html", "output_format": "Output all qualifying counties as a single line, sorted alphabetically by county name. Each county is formatted as County|SeniorPop|AmbulatoryDifficulty65Plus|AmbDiffPct|SeniorLivingAloneSharePct|NoVehiclePct|RenterNoVehiclePct|MobileHomePct|MedianIncome|BarrierPath, with counties separated by commas. For BarrierPath, use only ALONE, MOBILE_HOME, NO_VEHICLE, and RENTER_NO_VEHICLE, joined by + in that fixed order.", "oracle_answer": "Blair County|23013|5576|24.230|14.917|9.440|24.316|6.388|44033|ALONE+MOBILE_HOME, Cambria County|27058|6308|23.313|15.243|10.309|26.627|4.070|42917|ALONE+MOBILE_HOME+RENTER_NO_VEHICLE, Fayette County|25186|5886|23.370|14.961|9.302|22.290|10.489|40511|ALONE+MOBILE_HOME, Lycoming County|19486|4311|22.124|12.164|9.257|23.312|6.726|48731|ALONE+MOBILE_HOME, Mercer County|21537|4799|22.283|14.090|9.572|24.413|7.912|45831|ALONE+MOBILE_HOME, Philadelphia County|188185|56935|30.255|11.553|30.954|44.924|0.250|39770|NO_VEHICLE+RENTER_NO_VEHICLE, Schuylkill County|26613|6465|24.293|14.357|9.676|23.652|4.080|46573|ALONE+MOBILE_HOME", "metadata": {"State-Gated Retrieval": ["Use only Pennsylvania county data from the 2012–2016 American Community Survey 5-year estimates, and use the Pennsylvania statewide values from the same release as the comparison baseline.", "Retain only counties with a 65+ population of at least 15,000, a 65+ ambulatory-difficulty population of at least 4,000 with a share above the statewide share, and a median household income below the statewide median.", "The 65+ ambulatory-difficulty numerator must combine the four target cells from B18105 for ages 65–74 and 75+, male and female. The final county must also have at least two of the four barriers (Alone, No Vehicle, Renter No Vehicle, Mobile Home) above the statewide level."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["ACS tables B01001 and B18105 yield county senior-population totals and the full 65+ ambulatory-difficulty numerator", "statewide comparison tables yield Pennsylvania baselines plus county barrier metrics for living alone, no vehicle, renter no vehicle, mobile home, and median income", "final rows are produced only after the county candidate set is defined from the full ambulatory-difficulty numerator and the downstream six-table filters are applied"], "control_dependency": ["the ambulatory-difficulty numerator must include the full 65+ population across 65–74 and 75+ and male and female cells rather than only the 75+ cells", "downstream county eligibility and barrier labels must be determined from the county candidate set defined under that full ambulatory numerator", "the at-least-two-barriers test must compare every county metric against the same Pennsylvania statewide baselines from the same ACS release"], "freeze": {"historical_window": "2012–2016 American Community Survey 5-year estimates for Pennsylvania counties, with Pennsylvania statewide baselines taken from the same release"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Use only Pennsylvania county data from the 2012–2016 American Community Survey 5-year estimates, and use the Pennsylvania statewide values from the same release as the comparison baseline.", "Retain only counties with a 65+ population of at least 15,000, a 65+ ambulatory-difficulty population of at least 4,000 with a share above the statewide share, and a median household income below the statewide median.", "The 65+ ambulatory-difficulty numerator must combine the four target cells from B18105 for ages 65–74 and 75+, male and female. The final county must also have at least two of the four barriers (Alone, No Vehicle, Renter No Vehicle, Mobile Home) above the statewide level."], "exclusion_conditions": ["Exclude results that take only the 75-and-over ambulatory-difficulty cells and mistakenly treat them as the full 65+ count.", "Exclude results that do not use the Pennsylvania statewide baselines from the same release.", "Exclude counties whose final BarrierPath does not satisfy at least two barriers above the statewide level."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["County", "SeniorPop", "AmbulatoryDifficulty65Plus", "AmbDiffPct", "SeniorLivingAloneSharePct", "NoVehiclePct", "RenterNoVehiclePct", "MobileHomePct", "MedianIncome", "BarrierPath"], "county": "use the county display name including the word `County` and excluding a trailing state suffix", "dedup_key": "county FIPS / GEOID", "percentage_fields": "all percentage fields are plain numbers with exactly three decimal places and no percent sign", "numeric_fields": "population counts and median income are emitted as plain integers without commas", "BarrierPath": "use only ALONE, MOBILE_HOME, NO_VEHICLE, and RENTER_NO_VEHICLE, and when multiple labels apply join them with `+` in the fixed order ALONE+MOBILE_HOME+NO_VEHICLE+RENTER_NO_VEHICLE", "sorting_or_selection": "county name alphabetical ascending", "stop_condition": ["all candidate counties are validated under the same geography and product settings", "the senior ambulatory-difficulty gate and all barrier-signal gates are checked before the final line is emitted"]}}}
{"task_id": "climategov_011-g", "domain": "NOAA_NCEI_CLIMATE_AT_A_GLANCE", "autonomy_type": "ordered table", "oracle_output_cardinality": 3, "instruction": "Identify South Texas climate divisions that were near-normal or wetter over two longer periods and still had light heating demand in March 1966. Restrict the search strictly to the four divisions South Central, Upper Coast, South, and Lower Valley, using the exact names and numbering order from the official Texas climate-division page fragment.\n\nRetain only divisions that meet all three hard climate criteria: (1) the 10-month Palmer Z-Index anomaly ending February 1966 must be at least -0.05; (2) the 24-month Palmer Z-Index anomaly ending December 1966 must be at least -0.10; and (3) the actual March 1966 monthly Heating Degree Days value must be below 200°Df.\n\nSort qualifying divisions as follows: primary key is the sum of the February 1966 10-month Z-index anomaly and the December 1966 24-month Z-index anomaly, in descending order; secondary key is the March 1966 HDD actual value, in ascending order; tertiary key is the Texas division number, in ascending order.\n\nFinally, output all matching divisions as a single comma-separated string. Each entry must follow the format \"<Texas division number>|<division name>|<raw February 1966 10-month Z-index anomaly text>|<raw December 1966 24-month Z-index anomaly text>|<raw March 1966 monthly HDD actual value text>|<moisture score with two decimal places>\", where the moisture score is the sum of the two Z-index anomalies. If no division qualifies, output NONE in uppercase.", "start_url": "https://www.ncei.noaa.gov/access/monitoring/climate-at-a-glance/national", "output_format": "Output all matching divisions as a single comma-separated string. Each entry must follow the format \"<Texas division number>|<division name>|<raw February 1966 10-month Z-index anomaly text>|<raw December 1966 24-month Z-index anomaly text>|<raw March 1966 monthly HDD actual value text>|<moisture score with two decimal places>\", where the moisture score is the sum of the two Z-index anomalies. If no division qualifies, output NONE in uppercase.", "oracle_answer": "7|South Central|0.72|0.35|194°Df|1.07,10|Lower Valley|0.28|0.67|106°Df|0.95,9|South|0.03|0.37|148°Df|0.40", "metadata": {"State-Gated Retrieval": ["Filter only within the four South Texas divisions defined on the Texas divisional page: South Central, Upper Coast, South, and Lower Valley.", "Retain divisions that simultaneously satisfy: 1966-02 10-month Palmer Z-Index anomaly >= -0.05; 1966-12 24-month anomaly >= -0.10; 1966-03 1-month Heating Degree Days Value < 200°Df.", "Sort by sum of the two Z anomalies descending, then by 1966-03 HDD Value ascending, then by division number ascending."], "dependency_type": "Data", "intra_chain": true, "inter_chain": true, "data_dependency": ["The official Texas climate-division reference fixes the four South Texas division IDs and names that must be tracked across all later pages.", "The 1966-02 10-month Palmer Z-Index anomaly page and the 1966-12 24-month Palmer Z-Index anomaly page jointly identify the four target South Texas divisions.", "The 1966-03 1-month Heating Degree Days Value page supplies the last filter; final rows are produced only after intersecting the three divisional screens and then sorting by summed Z-anomaly score, HDD value, and division number."], "freeze": {"historical_window": "Texas climate divisions South Central, Upper Coast, South, and Lower Valley; metrics are fixed to the 1966-02 10-month Palmer Z-Index anomaly, 1966-12 24-month anomaly, and 1966-03 1-month Heating Degree Days Value pages."}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Filter only within the four South Texas divisions defined on the Texas divisional page: South Central, Upper Coast, South, and Lower Valley.", "Retain divisions that simultaneously satisfy: 1966-02 10-month Palmer Z-Index anomaly >= -0.05; 1966-12 24-month anomaly >= -0.10; 1966-03 1-month Heating Degree Days Value < 200°Df.", "Sort by sum of the two Z anomalies descending, then by 1966-03 HDD Value ascending, then by division number ascending."], "exclusion_conditions": ["Exclude any Texas divisional results outside the four specified South Texas divisions.", "Exclude results that misread the 1966-03 HDD page filter field as Anomaly instead of Value.", "Exclude any division that fails any time window or threshold condition."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["texas_division_no", "division_name", "feb_1966_10mo_z_text", "dec_1966_24mo_z_text", "mar_1966_hdd_text", "moisture_score"], "texas_division_no": "Canonical output uses the short Texas climate-division number; the scorer may accept an equivalent four-digit form, but the emitted oracle uses the short number.", "raw_text_fields": "The two Z-index anomaly fields and the March 1966 HDD field preserve the raw displayed text; comparison may ignore optional trailing HDD unit text such as `°Df`.", "moisture_score": "Compute as February 1966 10-month Z-index anomaly plus December 1966 24-month Z-index anomaly, formatted with exactly two decimal places.", "sorting_or_selection": {"primary": "moisture_score descending", "secondary": "March 1966 HDD actual value ascending after numeric normalization", "tertiary": "Texas division number ascending"}, "eligibility_rule": "Each Texas division is counted once per requested window; if any required window value is missing, exclude the division.", "stop_condition": ["All candidate Texas divisions are checked under the same base period and anomaly windows.", "The unique ordered output string is resolved before emission."]}}}
{"task_id": "climategov_012-g", "domain": "NOAA_NCEI_CLIMATE_AT_A_GLANCE", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I am researching historical climate patterns in Texas and want to find similar historical years as references for shoulder-season operational planning. To do this, I need to identify specific Texas climate divisions in NOAA's Climate at a Glance database. The divisions I am looking for must have experienced a shift from a dry long-term moisture background back toward near-normal or wetter conditions in the early 20th century, and must have had light heating loads in two different shoulder-season windows.\n\nSpecifically, the selected divisions must satisfy all four of the following hard climate criteria: first, the 10-month Palmer Z-Index anomaly ending April 1907 must be no higher than -0.30; second, the 10-month Z-Index anomaly ending April 1923 must be no lower than -0.20; third, the 2-month Heating Degree Days (HDD) actual value ending May 1915 must be below 50°Df; and fourth, the 5-month HDD actual value ending October 1922 must be below 20°Df.\n\nAfter identifying the qualifying divisions, sort them according to the following rules. The primary sort key is the \"rebound score\", defined as the April 1923 10-month Z-Index anomaly minus the April 1907 10-month Z-Index anomaly, sorted in descending order. In case of a tie, place the division with the smaller October 1922 5-month HDD actual value first. If still tied, place the division with the smaller Texas division number first.\n\nFinally, output all matching divisions as a single comma-separated string. Each matching item must be formatted exactly as: \"<Texas division number>|<4-digit division ID>|<division name>|<raw April 1907 10-month Z-Index anomaly text>|<raw April 1923 10-month Z-Index anomaly text>|<raw May 1915 2-month HDD actual value text>|<raw October 1922 5-month HDD actual value text>|<rebound score rounded to 2 decimal places>\". If after checking all divisions none meet the combined criteria, output NONE in uppercase.", "start_url": "https://www.ncei.noaa.gov/access/monitoring/climate-at-a-glance/national", "output_format": "Output all matching divisions as a single comma-separated string. Each matching item must be formatted exactly as: \"<Texas division number>|<4-digit division ID>|<division name>|<raw April 1907 10-month Z-Index anomaly text>|<raw April 1923 10-month Z-Index anomaly text>|<raw May 1915 2-month HDD actual value text>|<raw October 1922 5-month HDD actual value text>|<rebound score rounded to 2 decimal places>\". If after checking all divisions none meet the combined criteria, output NONE in uppercase.", "oracle_answer": "10|4110|Lower Valley|-0.36|1.46|10°Df|7°Df|1.82,9|4109|South|-0.33|0.53|23°Df|16°Df|0.86\n10|4110|Lower Valley|-0.36|1.46|10|7|1.82,9|4109|South|-0.33|0.53|23|16|0.86", "metadata": {"State-Gated Retrieval": ["Retain only Texas divisions that satisfy all four historical conditions: the 10-month Palmer Z-Index anomaly for April 1907 must be no higher than -0.30; the 10-month anomaly for April 1923 must be no lower than -0.20; the 2-month HDD Value for May 1915 must be less than 50°Df; and the 5-month HDD Value for October 1922 must be less than 20°Df.", "The rebound score is defined as the April 1923 10-month anomaly minus the April 1907 10-month anomaly.", "Sort first by rebound score descending, then by the October 1922 5-month HDD Value ascending, then by Texas division number ascending."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["Texas divisional pages yield the 1907-04 and 1923-04 10-month Palmer Z-Index anomalies plus the 1915-05 2-month and 1922-10 5-month HDD values for each division.", "The rebound score and the two shoulder-season HDD values are computed only after the same set of candidate divisions survives all four historical filters.", "The final output is produced after sorting the surviving divisions by rebound score, then 1922-10 HDD value, then division number."], "control_dependency": ["The workflow must cover the 1907, 1923, and 1915 pages as well as the 1922-10 5-month HDD page.", "Both HDD pages must be read from Value rather than Anomaly.", "The final rebound ranking must be determined after the 1922-10 5-month HDD page excludes 7 South Central and 8 Upper Coast, using ReboundScore as the ordering basis."], "freeze": {"historical_window": "Texas climate divisions evaluated on the 1907-04 and 1923-04 10-month Palmer Z-Index anomaly pages, the 1915-05 2-month HDD Value page, and the 1922-10 5-month HDD Value page."}, "answer_type": "mixed multi-line output"}, "rubric": {"inclusion_conditions": ["Retain only Texas divisions that satisfy all four historical conditions: the 10-month Palmer Z-Index anomaly for April 1907 must be no higher than -0.30; the 10-month anomaly for April 1923 must be no lower than -0.20; the 2-month HDD Value for May 1915 must be less than 50°Df; and the 5-month HDD Value for October 1922 must be less than 20°Df.", "The rebound score is defined as the April 1923 10-month anomaly minus the April 1907 10-month anomaly.", "Sort first by rebound score descending, then by the October 1922 5-month HDD Value ascending, then by Texas division number ascending."], "exclusion_conditions": ["Exclude results that stop after filtering only the 1907, 1923, and 1915 pages without also reading the 1922-10 5-month HDD page.", "Exclude results that mistakenly read the 1915 or 1922 HDD page as Anomaly instead of Value.", "Exclude any division that does not satisfy any one of the historical window conditions."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["texas_division_no", "division_id_4digit", "division_name", "apr_1907_10mo_z_text", "apr_1923_10mo_z_text", "may_1915_2mo_hdd_text", "oct_1922_5mo_hdd_text", "rebound_score"], "texas_division_no": "Emit the short Texas climate-division number in field 1 and the four-digit division ID in field 2.", "raw_text_fields": "The two Z-index anomaly fields preserve the raw displayed text; the two HDD fields preserve the raw displayed text, but comparison may ignore an optional trailing °Df suffix.", "rebound_score": "Compute as the April 1923 10-month Z-index anomaly minus the April 1907 10-month Z-index anomaly, formatted with exactly two decimal places.", "sorting_or_selection": {"primary": "rebound_score descending", "secondary": "October 1922 HDD actual value ascending after numeric normalization", "tertiary": "Texas division number ascending"}, "answer_set_note": "The current oracle stores two equivalent answer sets: one preserves the HDD unit suffixes and one strips them; treat the two HDD columns as equivalent after optional °Df removal.", "stop_condition": ["All candidate Texas divisions are checked under the same anomaly and HDD windows.", "The unique ordered output string is resolved before emission."]}}}
{"task_id": "comptox_010-g", "domain": "COMPTOX_CHEMEXPO", "autonomy_type": "ordered table", "oracle_output_cardinality": 5, "instruction": "Among the following five substances found in records related to painted products in Europe—aniline, cobalt, chromium, 3,3'-dichlorobenzidine, and dibenz[a,h]anthracene—select those whose pages explicitly contain the record 'Arts and crafts/Office supplies - body paint - tattoo ink, detected, Europe'. For each qualifying substance, produce a consolidated list that integrates historical keyword status and the deepest-level use categories. First, for each substance, count how many of the following five dated historical keywords are present (presence only, ignoring any leading numbers): Substances in Products - Canada (4/2014), Indirect additives food contact (10/2018), OEHHA Proposition 65 (3/2019), WA Children's Safe Product Act (4/2020), OEHHA Proposition 65 (1/2023). Also record the earliest and latest year-month among the hits. Second, for each substance, retain one category that best reflects its current use. After excluding the use category to which the above-mentioned painted-products record belongs, establish for each substance a terminal-use profile along two dimensions: 'daily consumer goods' (with the broad-category selection priority Formulation → Article → Occupation) and 'production materials' (with the broad-category selection priority Occupation → Article → Formulation). Finally, sort all substances by historical keyword hit count in descending order; for ties, sort by the number of products at the selected consumer-side node in descending order.", "start_url": "https://comptox.epa.gov/chemexpo/", "output_format": "Output format: <DTXSID>|<hit_count>|<earliest>|<latest>|<consumer_kind>|<consumer_general_puc_id>|<consumer_terminal_puc_id>|<consumer_level>|<consumer_products>|<material_kind>|<material_general_puc_id>|<material_terminal_puc_id>|<material_level>|<material_products>. Results separated by commas; if no qualifying items exist, output NONE.", "oracle_answer": "DTXSID8020090|5|2014-04|2023-01|Formulation|272|295|type|4|Occupation|301|356|family|27,DTXSID1031040|3|2018-10|2023-01|Formulation|77|94|type|7|Occupation|301|400|family|308,DTXSID9020409|2|2019-03|2023-01|Occupation|321|321|general|11|Occupation|321|321|general|11,DTXSID6020432|2|2019-03|2023-01|Article|313|313|general|1|Occupation|301|356|family|27,DTXSID3031022|1|2023-01|2023-01|Formulation|77|95|type|4|Occupation|301|356|family|185", "metadata": {"State-Gated Retrieval": ["Screen only the five specified substances, and the target substance page must explicitly contain the record 'Arts and crafts/Office supplies - body paint - tattoo ink, detected, Europe'.", "Count whether each of the five specified dated keywords appears, and record the earliest and latest year-month among the hits.", "Then independently select one consumer-side route and one production-material-side route, the former following Formulation → Article → Occupation and the latter following Occupation → Article → Formulation, and refine each route separately to its deepest level."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["each chemical page yields the five dated historical keywords, the Europe body-paint/tattoo-ink qualifying record, and the earliest/latest dated-hit months", "PUC category paths yield one consumer-side route chosen by Formulation -> Article -> Occupation and one source-side route chosen independently by Occupation -> Article -> Formulation", "final rows are produced only after the consumer-side and source-side routes are selected and refined separately for each chemical"], "control_dependency": ["the source-side route must be selected independently from the consumer-side route", "for each chemical, the second route refinement must follow the production-material priority order rather than the daily-consumer route", "family/type/general outputs must be determined only after both route selections are made independently"], "freeze": {"historical_window": "the five named substances as objectively existing chemicals, using their ChemExpo chemical pages, PUC category paths, keyword-state evidence, and Europe body-paint/tattoo-ink presence as relatively stable classification signals"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Screen only the five specified substances, and the target substance page must explicitly contain the record 'Arts and crafts/Office supplies - body paint - tattoo ink, detected, Europe'.", "Count whether each of the five specified dated keywords appears, and record the earliest and latest year-month among the hits.", "Then independently select one consumer-side route and one production-material-side route, the former following Formulation → Article → Occupation and the latter following Occupation → Article → Formulation, and refine each route separately to its deepest level."], "exclusion_conditions": ["Exclude substances that do not contain the specified Europe painted-products/tattoo-ink record.", "Exclude results that directly reuse the same consumer-side route for the production-material side.", "Exclude results that do not independently select and refine both routes according to the specified priority orders."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["DTXSID", "hit_count", "earliest", "latest", "consumer_kind", "consumer_general_puc_id", "consumer_terminal_puc_id", "consumer_level", "consumer_products", "material_kind", "material_general_puc_id", "material_terminal_puc_id", "material_level", "material_products"], "dated_hits": "count only the exact dated keyword strings defined in the source rubric", "date_format": "YYYY-MM", "kind": "the kind fields use the canonical labels Article, Formulation, or Occupation", "level": "the level fields use the canonical labels general, family, or type", "products_source": "the Products counts come from the local chemical-detail node for the selected branch, not from an overall aggregate shown elsewhere on the site", "sorting_or_selection": {"primary": "hit_count descending", "secondary": "consumer_products descending", "tertiary": "DTXSID ascending"}}}}
{"task_id": "comptox_011-g", "domain": "COMPTOX_CHEMEXPO", "autonomy_type": "ordered table", "oracle_output_cardinality": 6, "instruction": "In ChemExpo, for each of the six substances, determine whether it appears more often in material-contact uses or manufacturing-source uses, then select the most representative use category from the corresponding direction. Specifically, the scope is limited to these six substances: Di(2-Ethylhexyl) Phthalate, Melamine, Tris(2-Ethylhexyl) Phosphate, 4,5-Dihydro-2-Mercaptoimidazole, Phthalic Acid, and Triphenyl Phosphates Tert-Butylated. For each substance, three sets of data must be checked: first, a 'material-contact score' obtained by summing the occurrences of five specific labels (Indirect additives food contact (10/2018), Europe, Food contact items, Food contact items, Japan, entries containing 'Toys and children's products', and entries containing 'artificial_saliva' or 'artificial_sweat'); second, a 'manufacturing-source score' obtained by summing the occurrences of two specific labels (drinking_water, Europe, manufacturing, plastic_additive and Europe, plastic_additive); third, a regulatory profile generated by counting the total hits of four date-keyword labels (Indirect additives food contact (10/2018), OEHHA Proposition 65 (3/2019), WA Children's Safe Product Act (4/2020), OEHHA Proposition 65 (1/2023)) and extracting the earliest and latest month-year range spanned. Based on this, if the material-contact score is strictly greater than the manufacturing-source score, the substance is considered to favor material contact, and the most representative category is selected in the order Article → Formulation → Occupation; otherwise, it is considered to favor manufacturing source, and the selection order is Occupation → Article → Formulation. After determining the category, at each level, follow only the subcategory with the highest product count, continuing until no further subdivision is possible; if there is a tie at the same level, select the one with the smaller PUC ID.", "start_url": "https://comptox.epa.gov/chemexpo/", "output_format": "Finally, sort the substances by datedHits in descending order; for ties, sort by Products of the selected node in descending order. Output format: <DTXSID>|<contactHits>|<sourceHits>|<datedHits>|<earliest>|<latest>|<prioritySide>|<PUCkind><generalPUCID>|<terminalPUCID>|<level>|<Products>|<PUCDocuments>|<PUCCuratedChemicals>. Results are separated by commas; if datedHits is 0, write NA for both earliest and latest.", "oracle_answer": "DTXSID5020607|5|2|4|2018-10|2023-01|contact|Article|309|493|family|1|23|58,DTXSID5020601|4|1|3|2018-10|2023-01|contact|Article|309|309|general|1|5836|1827,DTXSID6020802|4|2|1|2018-10|2018-10|contact|Article|313|313|general|15|1104|675,DTXSID8021484|3|1|1|2018-10|2018-10|contact|Occupation|321|321|general|8|28313|4200,DTXSID0021414|5|0|1|2018-10|2018-10|contact|Article|309|309|general|3|5836|1827,DTXSID20872645|0|1|0|NA|NA|source|Occupation|318|381|family|5|274|107", "metadata": {"State-Gated Retrieval": ["Only consider the six specified substances, and first compute contactHits, sourceHits, and datedHits on the chemical page.", "prioritySide must be determined by these historical signals first, then select the most representative current use category along the corresponding direction.", "The current route selection must continue drilling down according to the given kind priority and the same-level smaller PUC ID rule."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["each chemical page yields contactHits, sourceHits, datedHits, and the earliest/latest dated-hit months from the historical keyword set", "after the priority side is chosen, the corresponding current PUC category path yields the representative kind, PUC path, product counts, and curated-chemical counts", "final rows are produced only after the historical hit scores determine the side first and the current route is then chosen with the side-specific priority and tie-break rules"], "control_dependency": ["historical contact/source signal scores must take priority over current maximum-Products paths", "the priority side must be determined from contactHits, sourceHits, and datedHits rather than from current visibility alone", "PUC path tie-breaks are valid only after the correct side has been fixed and the route has been traced under that side's rules"], "freeze": {"historical_window": "the six named substances as objectively existing chemicals, using their ChemExpo chemical pages, PUC category paths, historical keyword families, and keyword-hit evidence as relatively stable classification signals"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Only consider the six specified substances, and first compute contactHits, sourceHits, and datedHits on the chemical page.", "prioritySide must be determined by these historical signals first, then select the most representative current use category along the corresponding direction.", "The current route selection must continue drilling down according to the given kind priority and the same-level smaller PUC ID rule."], "exclusion_conditions": ["Exclude results that ignore historical signals and directly select the path with the maximum currently visible Products on the chemical page.", "Exclude results that start PUC route selection without first determining the contact/source direction.", "Exclude results where same-level tie-breaks do not follow the smaller PUC ID rule."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["DTXSID", "contactHits", "sourceHits", "datedHits", "earliest", "latest", "prioritySide", "PUCkind", "generalPUCID", "terminalPUCID", "level", "Products", "PUCDocuments", "PUCCuratedChemicals"], "dated_hits": "count only the exact dated keyword strings defined in the source rubric", "date_format": "YYYY-MM; if datedHits is 0, earliest and latest are both `NA`", "prioritySide": "emit `contact` when contactHits > sourceHits; otherwise emit `source`", "kind": "the PUCkind field uses the canonical labels Article, Formulation, or Occupation", "level": "the level field uses the canonical labels general, family, or type", "products_source": "Products comes from the selected terminal chemical-detail node", "terminal_statistics": "PUCDocuments and PUCCuratedChemicals come from the selected terminal PUC detail page Additional Statistics block", "sorting_or_selection": {"primary": "datedHits descending", "secondary": "Products descending", "tertiary": "DTXSID ascending"}}}}
{"task_id": "comptox_012-g", "domain": "COMPTOX_CHEMEXPO", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I want to first screen for Formulation categories with many products in ChemExpo, and then from each category pick one specific product type that carries a child or aerosol label and is most worth prioritizing. Specifically, I am looking for a set of results: for each Formulation category with a total product count of at least 15,000, find one optimal specific product type. This target type must belong to the first family under that category's default display order that meets the following conditions: the family has a total product count of at least 3,000 and no assumed attributes, and it must have product types whose allowed attributes include 'aerosol' and 'child', with those types themselves having no assumed attributes. Then, among all qualifying 'aerosol' and 'child' types under that family, select the one with the highest product share (type product count divided by family cumulative product count). In case of a tie in share, prioritize in order: higher Product Count, more Curated Chemicals, or smaller PUC ID, until a single winner is determined. Finally, retain one final result: prefer the specific product type with the higher product share; if shares are equal, compare Product Count, Curated Chemicals, and PUC ID in that order, retaining the type with the higher count, more chemicals, or smaller ID, until a single final type is obtained.", "start_url": "https://comptox.epa.gov/chemexpo/visualizations/", "output_format": "Finally, output results sorted by final share descending, in the format: <general category PUC ID>|<family PUC ID>|<specific product type PUC ID>|<side>|<share>|<Curated Chemicals>. Join results with commas and no spaces. If no qualifying results exist, output NONE.", "oracle_answer": "48|42|62|child|2160/3531|583,137|197|210|child|6934/16904|1135", "metadata": {"State-Gated Retrieval": ["First, retain only Formulation categories with a total product count of at least 15,000.", "In each candidate family, both an aerosol-side type and a child-side type must be found: the family itself must have no assumed attributes and a total product count of at least 3,000; both types must also have no assumed attributes.", "All selections within a family and between the two sides must follow the specified tie-break rules: share, Product Count, Curated Chemicals, PUC ID."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["The Formulation summary yields the high-product candidate families whose total product count is at least 15,000.", "Within each family, lower-level type pages yield the aerosol and child candidates, their assumed/allowed attributes, product counts, shares, and curated-chemical counts.", "Final rows are produced only after a family survives both aerosol and child branches and the winning type is selected by the layered share/product/chemical/PUC tie-break rules."], "control_dependency": ["Both the aerosol and child branches within the same family must be fully checked before that family can be retained.", "If one side is missing, the workflow must reject the family and continue to the next sibling family rather than stop at the first partial hit.", "The final winner can be selected only after both sides have been independently optimized within the family and then compared against each other."], "freeze": {"historical_window": "ChemExpo Formulation families as objectively existing catalog families, especially large families with total product count >= 15,000, evaluated by family/type identity and product/chemical counts as relatively stable catalog descriptors."}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["First, retain only Formulation categories with a total product count of at least 15,000.", "In each candidate family, both an aerosol-side type and a child-side type must be found: the family itself must have no assumed attributes and a total product count of at least 3,000; both types must also have no assumed attributes.", "All selections within a family and between the two sides must follow the specified tie-break rules: share, Product Count, Curated Chemicals, PUC ID."], "exclusion_conditions": ["Exclude results that stop as soon as a usable type is found on either side, without requiring the same family to cover both aerosol and child sides.", "Exclude results that stop as soon as one side is missing instead of continuing to check sibling families at the same level.", "Exclude results that do not follow the specified layered tie-break rules to select the winning type within the family and as the final winner."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["generalPUCID", "familyPUCID", "typePUCID", "side", "share", "CuratedChemicals"], "dedup_key": "(generalPUCID, familyPUCID, typePUCID, side)", "side": "emit the lower-case branch label exactly as determined in the source workflow", "share": "canonical output uses the raw fraction `type Products / family Cumulative Products` rather than a decimal", "CuratedChemicals": "emit the integer count from the selected type-level node", "sorting_or_selection": {"primary": "share descending by numeric fraction value", "secondary": "type-level Products descending", "tertiary": "CuratedChemicals descending, then PUC IDs ascending as tie-breaks"}}}}
{"task_id": "comptox_013-g", "domain": "COMPTOX_CHEMEXPO", "autonomy_type": "ordered table", "oracle_output_cardinality": 5, "instruction": "In ChemExpo, for the five substances Di(2-Ethylhexyl) Phthalate, Melamine, Phthalic Acid, Tris(2-Ethylhexyl) Phosphate, and Triphenyl Phosphates Tert-Butylated, check whether their historical labels are consistent with their current uses. The scope is limited to these five substances. For each substance, examine three groups of data. First, the material-contact score: check whether each of the following five label types appears, scoring 1 point per type: Indirect additives food contact (10/2018), Europe, Food contact items, Food contact items, Japan, any keyword containing Toys and children's products, and any keyword containing artificial_saliva or artificial_sweat. Second, the production-source score: check whether each of the following two label types appears, scoring 1 point per type: drinking_water, Europe, manufacturing, plastic_additive and Europe, plastic_additive. Third, dated-keyword hits: check only whether each of the following four keywords appears: Indirect additives food contact (10/2018), OEHHA Proposition 65 (3/2019), WA Children's Safe Product Act (4/2020), OEHHA Proposition 65 (1/2023); record the number of hits and the earliest and latest year-month among the hits. Then, if the material-contact score is strictly greater than the production-source score, treat it as leaning toward material contact and select a historically recommended category following the order Article → Formulation → Occupation; otherwise treat it as leaning toward production source and select following the order Occupation → Article → Formulation. After determining the category, follow the subcategory with the most products downward until no further subdivision is possible; if there is a tie at the same level, take the one with the smaller PUC ID. Separately, also find the current dominant category for this substance, ignoring historical clues, directly finding the path with the most products among all use categories, again following downward until no further subdivision is possible, breaking ties by smaller PUC ID. Finally, compare the two results: if the current dominant category is Article or Formulation and the historical clue also points to material contact, record ALIGNED; if the current dominant category is Occupation and the historical clue also points to production source, record ALIGNED; otherwise record MISALIGNED.", "start_url": "https://comptox.epa.gov/chemexpo/", "output_format": "Output MISALIGNED records first; within the same group, sort by dated hits descending; if hits are equal, sort by current dominant category product count descending. Format: <DTXSID>|<materialHits>|<sourceHits>|<datedHits>|<earliest>|<latest>|<historySide>|<recommendedKind>|<recommendedGeneral>|<recommendedTerminal>|<recommendedLevel>|<recommendedProducts>|<currentKind>|<currentGeneral>|<currentTerminal>|<currentLevel>|<currentProducts>|<status>. Separate records with commas. If dated hits is 0, write NA for both earliest and latest.", "oracle_answer": "DTXSID6020802|4|2|1|2018-10|2018-10|material|Article|Furniture and furnishings|Furniture and furnishings|general|15|Occupation|Raw materials|fireproof coatings|type|24|MISALIGNED,DTXSID8021484|3|1|1|2018-10|2018-10|material|Occupation|Laboratory supplies|Laboratory supplies|general|8|Occupation|Laboratory supplies|Laboratory supplies|general|8|MISALIGNED,DTXSID5020607|5|2|4|2018-10|2023-01|material|Article|Construction and building materials|flooring|family|1|Formulation|Home maintenance|paint|type|112|ALIGNED,DTXSID0021414|5|0|1|2018-10|2018-10|material|Article|Construction and building materials|Construction and building materials|general|3|Article|Construction and building materials|Construction and building materials|general|3|ALIGNED,DTXSID20872645|0|1|0|NA|NA|source|Occupation|Industrial products|hydraulic fluid|family|5|Occupation|Industrial products|hydraulic fluid|family|5|ALIGNED", "metadata": {"State-Gated Retrieval": ["Work only on the five specified substances, and first compute materialHits, sourceHits, and datedHits from the keyword-set blocks.", "The historically recommended category must be derived independently from these historical signals; the current dominant category must be derived separately from the current use path.", "Finally, compare only whether the broad direction of the historically recommended category and the current dominant category match, and mark ALIGNED or MISALIGNED accordingly."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["each chemical page yields materialHits, sourceHits, datedHits, and the historical recommendation side implied by the keyword set", "an independent current PUC category path yields the current dominant category, path labels, and classification counts for the same chemical", "final rows are produced only after the historical recommendation and current dominant category are computed independently and then compared at the coarse side level"], "control_dependency": ["the historical recommendation must be determined from the keyword-set evidence rather than from the current largest category", "historical recommendation must be computed from the keyword-set evidence before alignment is judged", "ALIGNED / MISALIGNED status is valid only after the historical and current category directions have been derived independently and then compared"], "freeze": {"historical_window": "the five named substances as objectively existing chemicals, using their ChemExpo chemical pages, PUC category paths, and specified keyword-set evidence as relatively stable recommendation signals"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Work only on the five specified substances, and first compute materialHits, sourceHits, and datedHits from the keyword-set blocks.", "The historically recommended category must be derived independently from these historical signals; the current dominant category must be derived separately from the current use path.", "Finally, compare only whether the broad direction of the historically recommended category and the current dominant category match, and mark ALIGNED or MISALIGNED accordingly."], "exclusion_conditions": ["Exclude results that only look at the current largest category and directly treat it as the historically recommended category.", "Exclude results that default to ALIGNED for all cases without independently deriving the historically recommended category.", "Exclude results that do not compute the historical and current paths separately but mix signals in the same route."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["DTXSID", "materialHits", "sourceHits", "datedHits", "earliest", "latest", "historySide", "recommendedKind", "recommendedGeneral", "recommendedTerminal", "recommendedLevel", "recommendedProducts", "currentKind", "currentGeneral", "currentTerminal", "currentLevel", "currentProducts", "status"], "dated_hits": "count only the exact dated keyword strings defined in the source rubric", "date_format": "YYYY-MM; if datedHits is 0, earliest and latest are both `NA`", "historySide": "emit the lower-case historical winner `material` or `source`", "recommended_fields": "these fields describe the historically suggested path", "current_fields": "these fields describe the current dominant path", "kind": "the kind fields use the canonical labels Article, Formulation, or Occupation", "level": "the level fields use the canonical labels general, family, or type", "status": "emit `ALIGNED` or `MISALIGNED` exactly", "sorting_or_selection": {"primary": "status with MISALIGNED before ALIGNED", "secondary": "datedHits descending", "tertiary": "currentProducts descending", "quaternary": "DTXSID ascending"}}}}
{"task_id": "consumerfinance_008-g", "domain": "CFPB_REPORTS", "autonomy_type": "ordered table", "oracle_output_cardinality": 3, "instruction": "For CFPB administrative enforcement actions filed between 2022-01-01 and 2024-12-31, produce a core-amount extraction report covering three product categories: Deposits, Furnishing, and Prepaid. For each category, identify the first action in the website’s default display order that satisfies all of the following: the action detail page provides both a Consent Order and a press release; the Products label corresponds exclusively and exactly to the target category; and the page explicitly states a consumer redress amount exceeding $5,000,000 along with a separate civil money penalty. After locating these three target actions, extract from each action’s Consent Order PDF the first occurrence of the consumer redress amount and the civil money penalty amount in the body text, along with their physical page numbers (PDF page 1 is p1).", "start_url": "https://www.consumerfinance.gov/enforcement/actions/", "output_format": "Output the results strictly in alphabetical order of the product category’s English name, using the format <Product>|<Respondent>|<Docket>|<FilingDate>|<RedressUSD>|p<RedressPage>|<PenaltyUSD>|p<PenaltyPage>. Join entries with commas and no spaces. If no qualifying case is found for a category, output <Product>|NONE.", "oracle_answer": "Deposits|Bank of America, N.A.|2023-CFPB-0006|2023-07-11|$80,400,000|p11|$60,000,000|p14,Furnishing|Hyundai Capital America|2022-CFPB-0005|2022-07-26|$13,200,000|p31|$6,000,000|p35,Prepaid|U.S. Bank National Association|2023-CFPB-0019|2023-12-19|$5,700,000|p22|$15,000,000|p26", "metadata": {"State-Gated Retrieval": ["Consider only administrative enforcement actions on the CFPB Enforcement actions page whose Initial filing date falls between 2022-01-01 and 2024-12-31.", "For each of the three product categories (Deposits, Furnishing, Prepaid), search top-down in the current default order for the first action that simultaneously satisfies: the page provides a Consent Order and a Press release; the Products label corresponds one-to-one with the target category; and the action detail page explicitly states consumer redress > $5,000,000 and also includes a civil money penalty.", "Finally, extract from the correct Consent Order the first-occurrence amounts and page numbers for consumer redress and civil money penalty."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["Enforcement-action list pages yield product-branch candidate actions in current default order.", "Each action detail page and linked consent-order PDF yield product singularity, document-stack presence, redress and civil-money-penalty amounts, and first-occurrence pages.", "Final rows are emitted only after identifying, for each product, the first action that survives detail-page and PDF validation."], "control_dependency": ["List-level “has Consent Order + Press release” hits must be validated on the detail page for product singularity and money thresholds.", "Prepaid candidates force document-level disambiguation among Consent Order, Stipulation, and terminating-order PDFs.", "Per-product first-match selection must explicitly skip rejected candidates and validate downstream PDFs before it is finalized."], "freeze": {"historical_window": "Administrative enforcement actions with Initial filing date from 2022-01-01 through 2024-12-31; current default case ordering and currently linked action documents."}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Consider only administrative enforcement actions on the CFPB Enforcement actions page whose Initial filing date falls between 2022-01-01 and 2024-12-31.", "For each of the three product categories (Deposits, Furnishing, Prepaid), search top-down in the current default order for the first action that simultaneously satisfies: the page provides a Consent Order and a Press release; the Products label corresponds one-to-one with the target category; and the action detail page explicitly states consumer redress > $5,000,000 and also includes a civil money penalty.", "Finally, extract from the correct Consent Order the first-occurrence amounts and page numbers for consumer redress and civil money penalty."], "exclusion_conditions": ["Exclude cases where Products is not a single target category, or where the case appears relevant only at the list level but fails validation on the detail page.", "Exclude cases that do not clearly satisfy redress > $5M and a separate civil money penalty.", "Exclude cases that mistakenly use a Stipulation or Order Terminating the Consent Order in place of the required Consent Order."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["Product", "Respondent", "Docket", "FilingDate", "RedressUSD", "RedressPage", "PenaltyUSD", "PenaltyPage"], "product_order": "alphabetical by Product name", "fallback_row": "if a product has no qualifying case, emit `<Product>|NONE`", "date_format": "FilingDate uses YYYY-MM-DD", "money_fields": "canonical output preserves the dollar sign and commas from the oracle, but equivalent comparison may ignore `$` and commas and compare the numeric USD amount", "page_fields": "page fields use the canonical `pN` form", "document_scope": "extract both money amounts from the Consent Order PDF only; do not substitute a stipulation, termination order, or other companion document", "selection_rule": "for each product, keep the first action in the default filtered order that fully satisfies every gate in the source rubric"}}}
{"task_id": "consumerfinance_009-g", "domain": "CFPB_REPORTS", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I want to organize the annual household financial stability survey reports in CFPB Reports, whose latest version page links directly to the 2023 and 2022 editions. I am interested in topics that appeared in the 2022 edition but no longer appear in subsequent annual reports, excluding the topic specifically used to mark \"belongs to the same survey series.\" For each such topic, I need the following information: the earliest subsequent year in which it disappeared; the first report not belonging to this series among the results related to these topics from 2022 and earlier; the total number of related resources listed in that report, how many of them link back to the same survey series, how many of those linked reports have appendices with letter labels and at least three sections; and the title, publication date, and appendix letter range of the final selected report.", "start_url": "https://www.consumerfinance.gov/data-research/research-reports/", "output_format": "Output a single line per topic, sorted alphabetically by topic name: <topic>|<first-missing-year>|<pivot-report-or-NONE>|<pivot-date-or-NONE>|<direct-related-count>|<same-series-count>|<qualified-count>|<representative-report-or-NONE>|<publication-date-or-NONE>|<appendix-span-or-NONE>", "oracle_answer": "Saving|2023|Consumer Savings App Strategies and Savings Outcomes|2022-12-07|4|2|1|Emergency Savings and Financial Security: Insights from the Making Ends Meet Survey and Consumer Credit Panel|2022-03-23|A-D,Student loans|2023|Report of the CFPB Education Loan Ombudsman|2022-10-20|0|0|0|NONE|NONE|NONE", "metadata": {"State-Gated Retrieval": ["First, identify the CFPB household financial well-being annual report series whose latest version links directly to the 2023 and 2022 editions.", "Using the 2022 edition as the baseline, retain only topics that appear in 2022 but are no longer retained in subsequent annual reports, and exclude the topic specifically marked as \"belongs to the same survey series.\"", "For each topic, find the earliest non-series pivot report from 2022 or earlier, then count its directly listed related resources, how many link back to the same survey series, how many qualify under the appendix rule, and the appendix span of the final qualifying report."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["current CFPB Reports pages yield the 2024 -> 2023 -> 2022 annual-survey chain and the 2022 topic baseline", "report detail pages and report-search results yield first-disappearance year, pivot report title/date, related-resource counts, and links back to the annual-survey series", "final rows are produced only after separating the first jump target from later return-to-series counting"], "control_dependency": ["requiring the pivot report itself to link back to the same survey series incorrectly rules out valid topic rows", "title extraction must use the pivot detail page rather than search-result snippets", "related-resource expansion must continue even when the pivot report itself does not link back to the same survey series"], "freeze": {"historical_window": "currently linked 2024, 2023, and 2022 CFPB household financial well-being annual reports; topic disappearance is judged over that three-year chain, and pivot-report search is restricted to 2022-and-earlier reports"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["First, identify the CFPB household financial well-being annual report series whose latest version links directly to the 2023 and 2022 editions.", "Using the 2022 edition as the baseline, retain only topics that appear in 2022 but are no longer retained in subsequent annual reports, and exclude the topic specifically marked as \"belongs to the same survey series.\"", "For each topic, find the earliest non-series pivot report from 2022 or earlier, then count its directly listed related resources, how many link back to the same survey series, how many qualify under the appendix rule, and the appendix span of the final qualifying report."], "exclusion_conditions": ["Exclude topics that are still retained in subsequent annual reports.", "Exclude results that incorrectly treat \"the pivot report itself must link back to the same survey series\" as a hard condition for column 3.", "Exclude search results that directly belong to the same survey series or reports after 2022."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["topic", "first-missing-year", "pivot-report-or-NONE", "pivot-date-or-NONE", "direct-related-count", "same-series-count", "qualified-count", "representative-report-or-NONE", "publication-date-or-NONE", "appendix-span-or-NONE"], "topic_order": "alphabetical by topic name", "date_format": "pivot-date and publication-date use YYYY-MM-DD when present", "appendix_span": "canonical span uses compact labels such as `A-D`", "none_rule": "if qualified-count is 0, representative-report, publication-date, and appendix-span are all `NONE`"}}}
{"task_id": "consumerfinance_010-g", "domain": "CFPB_REPORTS", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I want to clearly organize the two waves of the Debt Collection Rule issued by the CFPB in 2020, including their respective proposal sources, delayed-effective-date arrangements, final effective dates, and the accompanying quick-reference PDF information on the implementation page. Specifically, I need the following: the exact issue date of each Final Rule; the originating proposed rule source and name, and the public comment deadline; whether the 2021 delayed-effective-date proposed rule applies to this batch of final rules, its public comment deadline, and the actual effective date that ultimately took effect; the quick-reference PDF file name on the official implementation page, and the first subheading in the body (excluding introductory or background paragraphs).", "start_url": "https://www.consumerfinance.gov/rules-policy/final-rules/", "output_format": "Output the two waves as a single-line string sorted by final-rule issue date from earliest to latest, identifying each wave by its final rule issue month (YYYY-MM):\n<wave>|<final-rule-date>|<proposal-chain>|<proposal-comment-closes>|<delay-proposal-applied>|<delay-proposal-comments-close>|<controlling-effective-date>|<quick-reference-pdf>|<summary-heading>", "oracle_answer": "2020-10|2020-10-30|2019-05-07 main|2019-09-18|YES|2021-05-19|2021-11-30|Executive summary of the October 2020 final rule|Coverage and Definitions,2020-12|2020-12-18|2019-05-07 main+2020-02-21 supplemental|2019-09-18+2020-08-04|YES|2021-05-19|2021-11-30|Executive summary of the December 2020 final rule|Validation Information Requirements and Disclosures", "metadata": {"State-Gated Retrieval": ["The two 2020 Debt Collection Practices (Regulation F) final rules that share the same name must be processed separately, not merged into one.", "For each wave, independently confirm: the issue date, the corresponding proposal chain and public comment deadlines, whether the 2021 delayed-effective-date proposed rule covers it and its comment deadline, the final effective date, the corresponding implementation quick-reference PDF, and the first subheading in that PDF after skipping the introduction.", "The detail page, proposal chain, and summary PDF for the October wave and the December wave must each be verified independently."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["final-rules index pages and rule-development pages yield the two 2020 Regulation F final-rule waves and their proposal branches", "each final-rule detail page, delayed-effective-date proposal page, and implementation quick-reference PDF yields the wave-specific dates, linked proposals, comment deadlines, and first summary heading", "final rows are produced only after the October and December waves are traced separately end to end"], "control_dependency": ["two same-title final rules force correction from name-based collapsing to issue-date branching", "proposal-chain validation must be done per wave; otherwise the later December chain contaminates the October row", "implementation PDFs must be matched back to the correct wave rather than accepting the latest quick-reference page"], "freeze": {"historical_window": "the 2020-10 and 2020-12 Debt Collection Rule final-rule waves, their linked 2020 proposal pages, the 2021 delayed-effective-date proposal, and the current implementation-page quick-reference PDFs"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["The two 2020 Debt Collection Practices (Regulation F) final rules that share the same name must be processed separately, not merged into one.", "For each wave, independently confirm: the issue date, the corresponding proposal chain and public comment deadlines, whether the 2021 delayed-effective-date proposed rule covers it and its comment deadline, the final effective date, the corresponding implementation quick-reference PDF, and the first subheading in that PDF after skipping the introduction.", "The detail page, proposal chain, and summary PDF for the October wave and the December wave must each be verified independently."], "exclusion_conditions": ["Exclude results that merge the two identically named 2020 final rules into one.", "Exclude results that project the December wave's proposal chain or executive-summary heading onto the October wave.", "Exclude results that are not independently verified against each wave's own final-rule detail page and corresponding PDF."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["wave", "final-rule-date", "proposal-chain", "proposal-comment-closes", "delay-proposal-applied", "delay-proposal-comments-close", "controlling-effective-date", "quick-reference-pdf", "summary-heading"], "wave": "wave is the final-rule issue month in YYYY-MM", "date_format": "all standalone date fields use YYYY-MM-DD", "proposal_chain": "canonical chain joins proposal milestones with `+`, using pieces such as `2019-05-07 main` or `2020-02-21 supplemental` in chronological order", "proposal_comment_closes": "join multiple comment-close dates with `+` in the same order as proposal_chain", "delay_proposal_applied": "emit `YES` when a delay proposal is part of the chain; otherwise emit `NO`", "quick_reference_pdf": "use the displayed quick-reference PDF title, not the filename", "summary_heading": "use the first substantive heading from the final-rule PDF", "sorting_or_selection": "final-rule-date ascending"}}}
{"task_id": "consumerfinance_011-g", "domain": "CFPB_REPORTS", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "On the CFPB enforcement actions page, find the most recent 2024 administrative enforcement action that has since been terminated, separately for Credit Cards and Deposits, and align the termination order with the corresponding clause information in the original Consent Order. Specifically, identify one action for each of the two Products labels: Credit Cards and Deposits. The candidate pool is limited to administrative enforcement actions under Enforcement actions with an Initial filing date between 2024-01-01 and 2024-12-31. For each product label, select the action with the latest Initial filing date among those meeting all of the following conditions (if tied, pick the one with the smaller docket number): Initial filing date between 2024-01-01 and 2024-12-31; case status changed to Expired, Terminated, or Dismissed; the document set includes a Consent Order, a Stipulation, an Order Terminating the Consent Order, and a press release; and the Products label currently maps uniquely to the target product. For each selected action, extract the following from its Order Terminating the Consent Order: TerminationDate (the termination date); TerminationAuthority (the exact text of the termination authority at its first occurrence); TermAuthorityPage (the PDF physical page number where this text first appears). Additionally, find the first explicit reference to a paragraph number of the original Consent Order within the termination order text; if it reads \"Paragraphs X and Y\", take only X. Then go back to the original Consent Order and find the section heading that contains that paragraph.", "start_url": "https://www.consumerfinance.gov/enforcement/actions/", "output_format": "Finally, output a single-line string in alphabetical order by product name: <Product>|<Respondent>|<Docket>|<InitialFilingDate>|<TerminationDate>|<TerminationAuthority>|p<TermAuthorityPage>|<ReferencedParagraph>|<ConsentSectionHeading>. If a product has no qualifying case, output: <Product>|NONE", "oracle_answer": "Credit Cards|Apple Inc.|2024-CFPB-0012|2024-10-23|2025-09-22|12 U.S.C. § 5563(b)(3)|p1|83|ADMINISTRATIVE PROVISIONS,Deposits|Navy Federal Credit Union|2024-CFPB-0014|2024-11-07|2025-07-01|12 U.S.C. § 5563(b)(3)|p1|106|ADMINISTRATIVE PROVISIONS", "metadata": {"State-Gated Retrieval": ["Consider only administrative enforcement actions with an Initial filing date in 2024, and search for cases separately under Credit Cards and Deposits.", "The case must have a status of Expired, Terminated, or Dismissed, and the page must simultaneously provide a Consent Order, Stipulation, Order Terminating the Consent Order, and Press release, with the Products label mapping one-to-one to the target category.", "For each category, retain only the most recent qualifying case determined by Initial filing date, and compile the corresponding clause information from the termination order and the original Consent Order."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["Enforcement-action list pages yield 2024 candidate actions by product branch.", "Each action detail page and linked Consent Order, Stipulation, Order Terminating the Consent Order, and Press release documents yield status, initial filing date, single-product validity, and clause references.", "Final rows are selected only after per-product comparison on Initial filing date and document-stack validation."], "control_dependency": ["\"Latest\" must be determined by Initial filing date rather than termination date.", "Product filters on the list page are insufficient until the action detail page confirms a single matching product.", "Clause extraction in the terminating order must stay aligned with the original Consent Order rather than a superficially similar section."], "freeze": {"historical_window": "Administrative enforcement actions with Initial filing date from 2024-01-01 through 2024-12-31; current action status and currently linked Consent Order, Stipulation, terminating-order, and Press release documents."}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Consider only administrative enforcement actions with an Initial filing date in 2024, and search for cases separately under Credit Cards and Deposits.", "The case must have a status of Expired, Terminated, or Dismissed, and the page must simultaneously provide a Consent Order, Stipulation, Order Terminating the Consent Order, and Press release, with the Products label mapping one-to-one to the target category.", "For each category, retain only the most recent qualifying case determined by Initial filing date, and compile the corresponding clause information from the termination order and the original Consent Order."], "exclusion_conditions": ["Exclude results that use termination date instead of Initial filing date to determine \"latest\".", "Exclude cases with an incomplete document stack or where Products does not map to a single target category.", "Exclude cases where the required clause correspondence between the terminating order and the original Consent Order cannot be established."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["Product", "Respondent", "Docket", "InitialFilingDate", "TerminationDate", "TerminationAuthority", "TermAuthorityPage", "ReferencedParagraph", "ConsentSectionHeading"], "product_order": "Alphabetical by Product name", "fallback_row": "If a product has no qualifying case, emit `<Product>|NONE`", "date_format": "InitialFilingDate and TerminationDate use YYYY-MM-DD", "termination_authority": "Use the canonical statute string extracted from the termination order", "page_fields": "TermAuthorityPage uses the canonical `pN` form", "ReferencedParagraph": "Emit the paragraph number as a plain numeric string", "ConsentSectionHeading": "Emit the uppercase heading of the consent-order section containing the referenced paragraph"}}}
{"task_id": "consumerfinance_012-g", "domain": "CFPB_COMPLAINT_DATABASE", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I am preparing an internal outreach memo on historical complaint clusters and want to track a specific set of complaints: in the CFPB Complaint Database, identify complaints under the current product bucket covering credit reporting and other personal consumer reports, received between 2018-01-01 and 2021-12-31, tagged as Servicemember, with a public narrative, and whose text explicitly mentions medical bills or doctors. Please provide two representative complaint clusters: the first is the peak month of this cohort; the second is the latest month in this cohort whose complaint count is still at least half of the peak count. For each cluster, report the total complaint count for that month, the state with the highest complaint count in that month and its count, the company with the highest complaint count in that state and month and its count, and the earliest complaint received by that company in that state and month, including its ID, date, Issue, and Sub-issue. Output a single line with Peak first and LatestHalfPeak second, using the format: <Cluster>|<Month>|<MonthComplaints>|<TopState>|<StateComplaints>|<TopCompany>|<CompanyComplaints>|<EarliestComplaintID>|<EarliestDate>|<Issue>|<SubIssue>.", "start_url": "https://www.consumerfinance.gov/data-research/consumer-complaints/", "output_format": "Output a single line with Peak first and LatestHalfPeak second, using the format: <Cluster>|<Month>|<MonthComplaints>|<TopState>|<StateComplaints>|<TopCompany>|<CompanyComplaints>|<EarliestComplaintID>|<EarliestDate>|<Issue>|<SubIssue>. Join the Peak and LatestHalfPeak clusters with a comma. For ties: months are ordered chronologically; states are ordered by state abbreviation alphabetically ascending, taking the first; companies are ordered by company name alphabetically ascending, taking the first; for the earliest complaint, if multiple complaints share the same date, take the one with the smallest Complaint ID.", "oracle_answer": "Peak|2018-02|46|CA|28|TRANSUNION INTERMEDIATE HOLDINGS, INC.|9|2806154|2018-02-07|Improper use of your report|Reporting company used your report improperly,LatestHalfPeak|2021-11|23|CA|12|EQUIFAX, INC.|3|4928768|2021-11-19|Problem with a credit reporting company's investigation into an existing problem|Their investigation did not fix an error on your report", "metadata": {"State-Gated Retrieval": ["Only consider complaints under the current product bucket for credit reporting / other personal consumer reports, received from 2018-01-01 to 2021-12-31, tagged as Servicemember, with a public narrative, and whose text explicitly mentions medical bills or doctors.", "The first cluster must be the peak month of this cohort; the second cluster must be the latest month in this cohort whose complaint count is at least half of the peak count.", "For each cluster, recompute the top company within the corresponding month-state subset, then locate the earliest complaint received by that company in that state and month."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["current complaint-form taxonomy yields the product bucket that contains credit reporting / other personal consumer reports", "filtered complaint records yield month counts, month-state subsets, top companies, and earliest complaint details inside each chosen subset", "final rows are produced only after peak month and latest-half-peak month are determined, and each downstream aggregate is computed inside its narrower month-state cohort"], "control_dependency": ["the latest month must be defined as the latest month whose count is at least ceil(PeakCount / 2)", "overall top-company counts cannot be projected into a month-state subset without recomputation", "earliest complaint selection is only valid after the month, state, and company filters are all frozen"], "freeze": {"historical_window": "Complaint Database records received from 2018-01-01 through 2021-12-31 under the current product-bucket mapping for the credit-reporting / other personal consumer reports family; Servicemember + public narrative + medical bills/doctors text cohort"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Only consider complaints under the current product bucket for credit reporting / other personal consumer reports, received from 2018-01-01 to 2021-12-31, tagged as Servicemember, with a public narrative, and whose text explicitly mentions medical bills or doctors.", "The first cluster must be the peak month of this cohort; the second cluster must be the latest month in this cohort whose complaint count is at least half of the peak count.", "For each cluster, recompute the top company within the corresponding month-state subset, then locate the earliest complaint received by that company in that state and month."], "exclusion_conditions": ["Exclude results that directly treat the latest month with any complaints as the second cluster.", "Exclude results that first find an overall top company in the full cohort and then project it onto a month-state subset.", "Exclude results that do not first derive the half-peak threshold from PeakCount."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["Cluster", "Month", "MonthComplaints", "TopState", "StateComplaints", "TopCompany", "CompanyComplaints", "EarliestComplaintID", "EarliestDate", "Issue", "SubIssue"], "cluster_order": ["Peak", "LatestHalfPeak"], "date_format": {"Month": "YYYY-MM", "EarliestDate": "YYYY-MM-DD"}, "TopState": "use the uppercase two-letter state abbreviation", "TopCompany": "use the company display name as shown in the source data", "tie_breaks": "for tied months use chronological order; for tied states use state-code alphabetical order; for tied companies use company-name alphabetical order; for the earliest complaint use EarliestDate ascending then Complaint ID ascending", "SubIssue": "emit `NONE` when the complaint row has no sub-issue value"}}}
{"task_id": "consumerfinance_013-g", "domain": "CFPB_COMPLAINT_DATABASE", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I am collecting data for an internal crypto-complaint concentration memo and need to identify two representative complaint-month clusters from the CFPB complaint database and extract their core data records.\nThe data must come from a specific cohort: the product category must correspond to the current form’s bucket for Virtual currency, with the sub-product explicitly set to Virtual currency; the time window is strictly 2019-08-26 through 2022-08-26.\nWithin this cohort, the two characteristic months to identify are:\n• Original peak month (AllCompanies): the month with the absolute highest complaint volume in the cohort without removing any company.\n• Residual peak month (ResidualAfterTop2): after identifying the two companies with the highest complaint volume in the original peak month and completely removing all records of those two companies from the entire three-year analysis period, the month with the highest complaint volume that emerges from the remaining data.\nFor these two specific months, I need detailed attributes: total complaints for that month, the names of the two removed companies and their combined complaint contribution in the original peak month, the state with the highest complaint volume in that month and its volume, and the earliest complaint detail (including ID, date, Issue, and Sub-issue) for the most complained-about company within that state and month.\nPlease confirm the identity of these two months and output one formatted string per month with no spaces, in the order AllCompanies first, ResidualAfterTop2 second. The field sequence per row is strictly: <Cluster>|<Month>|<MonthComplaints>|<ExcludedTop2Companies>|<ExcludedTop2PeakMonthComplaints>|<TopState>|<StateComplaints>|<TopCompany>|<CompanyComplaints>|<EarliestComplaintID>|<EarliestDate>|<Issue>|<SubIssue>\nSpecial notes:\n• The <Cluster> field must be filled with AllCompanies or ResidualAfterTop2 accordingly.\n• ExcludedTop2Companies and ExcludedTop2PeakMonthComplaints in both rows refer to the two companies identified from the original peak month and their combined complaint volume at that time.\n• The four Earliest-related fields must correspond to the earliest complaint record for the top company in that state and month.\n• To ensure a unique answer: output the two records in the order AllCompanies, ResidualAfterTop2; list ExcludedTop2Companies in descending order of their complaint volume in the original peak month, breaking ties alphabetically by field value; if the earliest complaint date is tied, take the one with the smallest Complaint ID.", "start_url": "https://www.consumerfinance.gov/data-research/consumer-complaints/", "output_format": "Output one formatted string per row with no spaces, in the order AllCompanies first, ResidualAfterTop2 second. The field sequence per row is strictly: <Cluster>|<Month>|<MonthComplaints>|<ExcludedTop2Companies>|<ExcludedTop2PeakMonthComplaints>|<TopState>|<StateComplaints>|<TopCompany>|<CompanyComplaints>|<EarliestComplaintID>|<EarliestDate>|<Issue>|<SubIssue>.", "oracle_answer": "AllCompanies|2021-05|336|Coinbase, Inc.+BAM Management US Holdings Inc.|226|CA|58|Coinbase, Inc.|21|4341451|2021-05-01|Money was not available when promised|NONE,ResidualAfterTop2|2020-06|180|Coinbase, Inc.+BAM Management US Holdings Inc.|226|OH|28|PNC Bank N.A.|22|3687376|2020-06-07|Other transaction problem|NONE", "metadata": {"State-Gated Retrieval": ["Only consider Complaint Database records under the current product bucket for Virtual currency, with sub-product Virtual currency, and received between 2019-08-26 and 2022-08-26.", "First determine the AllCompanies original peak month, then identify the top two companies by complaint volume in that peak month.", "These two companies must be completely removed from the entire three-year cohort, then recompute the residual peak month and all downstream fields based on the remaining cohort."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["current complaint-form taxonomy yields the product bucket for Virtual currency", "complaint records within the fixed three-year cohort yield the all-company peak month, per-month top companies, and residual-cohort recomputations after company removal", "final rows are produced only after the top-2 companies from the original peak month are removed from the full cohort and downstream aggregates are computed on the resulting cohort"], "control_dependency": ["peak-month top companies must be removed from the entire cohort, not only from the original peak month", "residual peak detection must be computed from the rewritten cohort rather than by local patching of the original month", "downstream state / company / earliest-complaint fields are valid only after the global company-removal step is locked"], "freeze": {"historical_window": "Complaint Database records received from 2019-08-26 through 2022-08-26 under the current product-bucket mapping for Virtual currency"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Only consider Complaint Database records under the current product bucket for Virtual currency, with sub-product Virtual currency, and received between 2019-08-26 and 2022-08-26.", "First determine the AllCompanies original peak month, then identify the top two companies by complaint volume in that peak month.", "These two companies must be completely removed from the entire three-year cohort, then recompute the residual peak month and all downstream fields based on the remaining cohort."], "exclusion_conditions": ["Exclude results that only locally delete the top-2 company records from the original peak month.", "Exclude results that do not reapply the top-2 company set to the entire cohort before recomputing the residual peak.", "Exclude downstream state, company, or earliest-complaint fields based on a cohort that has not been recomputed after removal."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["Cluster", "Month", "MonthComplaints", "ExcludedTop2Companies", "ExcludedTop2PeakMonthComplaints", "TopState", "StateComplaints", "TopCompany", "CompanyComplaints", "EarliestComplaintID", "EarliestDate", "Issue", "SubIssue"], "cluster_order": ["AllCompanies", "ResidualAfterTop2"], "date_format": {"Month": "YYYY-MM", "EarliestDate": "YYYY-MM-DD"}, "ExcludedTop2Companies": "join the two excluded company names with `+` in source-peak complaint-count descending order, breaking ties alphabetically", "TopState": "use the uppercase two-letter state abbreviation", "TopCompany": "use the company display name as shown in the source data", "tie_breaks": "for the earliest complaint use EarliestDate ascending then Complaint ID ascending", "SubIssue": "emit `NONE` when the complaint row has no sub-issue value"}}}
{"task_id": "consumerfinance_014-g", "domain": "CFPB_COMPLAINT_DATABASE", "autonomy_type": "ordered table", "oracle_output_cardinality": 4, "instruction": "I am writing an internal crypto-complaint sensitivity waterfall memo and need to extract four characteristic months (Clusters) at different filtering stages from the CFPB complaint database, along with their core data.\n\nThe underlying complaint cohort has clear boundaries: time is strictly limited to between 2020-01-01 and 2022-08-26, and it includes only complaint records classified under the \"Virtual currency\" subcategory within the current product taxonomy.\n\nBased on this underlying cohort, I need to identify four characteristic peak months representing different successive exclusion stages:\n• BaselinePeak: The absolute peak month with the highest total complaint volume in the initial complaint pool.\n• AfterDominantPublicResponse: After globally removing all complaints related to the dominant Company public response from the BaselinePeak month, the new peak month in the remaining data pool.\n• AfterDominantPublicResponseAndIssue: After further globally removing all complaints related to the dominant Issue from the second-round peak month, the third-round peak month formed from the remaining data.\n• AfterDominantPublicResponseIssueAndCompany: After finally globally removing all complaints related to the most complained-about Company from the third-round peak month, the final peak month that emerges from the remaining data.\n\nFor these four characteristic months, I need their data under each respective filtering state: total complaints for that month, the new exclusion dimension triggered in this round (NewExclusionType) and its specific value (NewExclusionValue), the frequency of that excluded value in its \"source peak month\", the state with the highest complaint volume in that month and its complaint count, the most complained-about company within that state and its complaint count, and the detailed information (ID, date, Issue, and Sub-issue) of the earliest complaint received by that top company in that state and characteristic month.\n\nPlease confirm the identity and data for these four characteristic months, and output four formatted strings with no spaces, strictly in the specific order: BaselinePeak, AfterDominantPublicResponse, AfterDominantPublicResponseAndIssue, AfterDominantPublicResponseIssueAndCompany. The field sequence for each row must be fixed as:\n<Cluster>|<Month>|<MonthComplaints>|<NewExclusionType>|<NewExclusionValue>|<NewExclusionCountInSourcePeak>|<TopState>|<StateComplaints>|<TopCompany>|<CompanyComplaints>|<EarliestComplaintID>|<EarliestDate>|<Issue>|<SubIssue>\n\nFor ties, always select in ascending lexicographic order; for the earliest complaint with the same date, select by ascending Complaint ID numerically. Normalize all empty values to NONE before sorting.", "start_url": "https://www.consumerfinance.gov/data-research/consumer-complaints/", "output_format": "Output four formatted strings with no spaces, strictly in the specific order: BaselinePeak, AfterDominantPublicResponse, AfterDominantPublicResponseAndIssue, AfterDominantPublicResponseIssueAndCompany. The field sequence for each row must be fixed as: <Cluster>|<Month>|<MonthComplaints>|<NewExclusionType>|<NewExclusionValue>|<NewExclusionCountInSourcePeak>|<TopState>|<StateComplaints>|<TopCompany>|<CompanyComplaints>|<EarliestComplaintID>|<EarliestDate>|<Issue>|<SubIssue>. For ties, always select in ascending lexicographic order; for the earliest complaint with the same date, select by ascending Complaint ID numerically. Normalize all empty values to NONE before sorting.", "oracle_answer": "BaselinePeak|2021-05|336|NONE|NONE|0|CA|58|Coinbase, Inc.|21|4341451|2021-05-01|Money was not available when promised|NONE,AfterDominantPublicResponse|2022-07|31|Company public response|NONE|321|CA|6|BANK OF AMERICA, NATIONAL ASSOCIATION|4|5735227|2022-07-05|Other transaction problem|NONE,AfterDominantPublicResponseAndIssue|2021-10|13|Issue|Fraud or scam|26|TX|2|BANK OF AMERICA, NATIONAL ASSOCIATION|1|4821611|2021-10-18|Other transaction problem|NONE,AfterDominantPublicResponseIssueAndCompany|2021-06|8|Company|Foris DAX, Inc.|5|NC|3|TRUIST FINANCIAL CORPORATION|2|4462160|2021-06-15|Money was not available when promised|NONE", "metadata": {"State-Gated Retrieval": ["Only consider Complaint Database records under the Virtual currency subcategory within the current product bucket, with a received date between 2020-01-01 and 2022-08-26.", "Sequentially determine the peak months for the four stages: BaselinePeak, AfterDominantPublicResponse, AfterDominantIssue, and AfterDominantCompany, as required by the task.", "Each identified dominant Company public response, dominant Issue, and dominant Company must be written back to the corresponding upstream residual cohort, globally removed, and then the next round of statistics must be recalculated."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["current complaint-form taxonomy yields the product bucket for Virtual currency", "complaint records yield the baseline peak plus the dominant Company public response, dominant Issue, and dominant Company at successive residual-peak states", "final rows are produced only after each dominant value is applied to the full upstream residual cohort and then used to determine the next peak"], "control_dependency": ["dominant values found in a peak month cannot be deleted only inside that month; each must be propagated to the full upstream cohort", "each new residual peak depends on the prior round's global rewrite, so local month-by-month deletion gives the wrong chain", "downstream counts and exemplar complaints are valid only after the full three-stage waterfall is executed"], "freeze": {"historical_window": "Complaint Database records received from 2020-01-01 through 2022-08-26 under the current product-bucket mapping for Virtual currency"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Only consider Complaint Database records under the Virtual currency subcategory within the current product bucket, with a received date between 2020-01-01 and 2022-08-26.", "Sequentially determine the peak months for the four stages: BaselinePeak, AfterDominantPublicResponse, AfterDominantIssue, and AfterDominantCompany, as required by the task.", "Each identified dominant Company public response, dominant Issue, and dominant Company must be written back to the corresponding upstream residual cohort, globally removed, and then the next round of statistics must be recalculated."], "exclusion_conditions": ["Exclude results that only delete records locally within the current peak month without writing back to the entire upstream cohort.", "Exclude results that do not re-find the next peak month on each round's residual cohort.", "Exclude intermediate or final fields based on downstream statistics that were not recalculated globally."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["Cluster", "Month", "MonthComplaints", "NewExclusionType", "NewExclusionValue", "NewExclusionCountInSourcePeak", "TopState", "StateComplaints", "TopCompany", "CompanyComplaints", "EarliestComplaintID", "EarliestDate", "Issue", "SubIssue"], "cluster_order": ["BaselinePeak", "AfterDominantPublicResponse", "AfterDominantPublicResponseAndIssue", "AfterDominantPublicResponseIssueAndCompany"], "date_format": {"Month": "YYYY-MM", "EarliestDate": "YYYY-MM-DD"}, "NewExclusionType": "emit the canonical exclusion-type label from the source workflow; when no new exclusion is applied, emit `NONE`", "NewExclusionValue": "emit the canonical excluded value; when no value exists, emit `NONE`", "TopState": "use the uppercase two-letter state abbreviation", "TopCompany": "use the company display name as shown in the source data", "tie_breaks": "normalize empty values to `NONE` before sorting; for other ties use lexicographic order ascending, and for the earliest complaint use EarliestDate ascending then Complaint ID ascending", "SubIssue": "emit `NONE` when the complaint row has no sub-issue value"}}}
{"task_id": "europemc_001-g", "domain": "EUROPEPMC_PMC", "autonomy_type": "ordered table", "oracle_output_cardinality": 21, "instruction": "I want to consolidate into a single line the admission variables actually used by several prediction schemes in this 2020 COVID-19 mortality risk prediction paper, so that it is easy to see which variables overlap across schemes and which appear only in the comparator schemes.\nSpecifically, the target paper is a 2020 open-access review whose title contains all three terms \"prognosis model\", \"mortality risk\", and \"COVID-19\", and whose main text discusses Wuhan inpatients.\nThe variables to retain fall into two categories: (1) admission variables that appear in both the Random Forest and XGBoost variable importance rankings; (2) admission variables that do not appear in both rankings but are additional variables present in the CURB-65 or NEWS schemes used by the authors as comparators.\nFormat each variable as follows:\nvariable~RF_rank_or_0~XGB_rank_or_0~F~C~N~X\nwhere:\n\t•\tRF is the variable's rank in the Random Forest ranking, or 0 if absent;\n\t•\tXGB is the variable's rank in the XGBoost ranking used in the paper, or 0 if absent;\n\t•\tF, C, N, X indicate, respectively, whether the variable entered the authors' final logistic regression model, belongs to CURB-65, belongs to NEWS, and belongs to the XGBoost model used as a comparator in the paper.\nFinally, sort all variables alphabetically by variable name, join them with | into a single line; if the target paper cannot be found, output NONE.", "start_url": "https://europepmc.org/advancesearch", "output_format": "Format each variable as follows:\nvariable~RF_rank_or_0~XGB_rank_or_0~F~C~N~X\nwhere:\n\t•\tRF is the variable's rank in the Random Forest ranking, or 0 if absent;\n\t•\tXGB is the variable's rank in the XGBoost ranking used in the paper, or 0 if absent;\n\t•\tF, C, N, X indicate, respectively, whether the variable entered the authors' final logistic regression model, belongs to CURB-65, belongs to NEWS, and belongs to the XGBoost model used as a comparator in the paper.\nFinally, sort all variables alphabetically by variable name, join them with | into a single line; if the target paper cannot be found, output NONE.", "oracle_answer": "Age~4~2~1~1~0~0|Alanine aminotransferase~17~11~0~0~0~0|Aspartate transaminase~6~4~0~0~0~0|Blood urea nitrogen~5~7~0~1~0~0|Confusion~0~0~0~1~1~0|CRP~2~6~1~0~0~1|Creatinine~9~13~0~0~0~0|Diastolic pressure~16~17~0~1~0~0|Heart rate~15~14~0~0~1~0|LDH~1~1~1~0~0~1|Lymphocyte count~3~5~0~0~0~1|Monocytes~12~16~0~0~0~0|Neutrophil~7~3~0~0~0~0|Normal platelet~10~9~0~0~0~0|Oxygen inhalation~0~0~0~0~1~0|Respiratory rate~11~10~0~1~1~0|SPO2~0~0~0~0~1~0|Systolic pressure~13~12~0~1~1~0|Temperature~18~18~0~0~1~0|Total bilirubin~14~15~0~0~0~0|White blood cell~8~8~0~0~0~0", "metadata": {"State-Gated Retrieval": ["First locate the target paper: a 2020 open-access review whose title contains all three terms \"prognosis model\", \"mortality risk\", and \"COVID-19\", and whose main text discusses Wuhan inpatients.", "Retain only two categories of variables: admission variables that appear in both the Random Forest and XGBoost rankings; or admission variables that, although not appearing in both rankings, belong to the CURB-65 or NEWS comparator schemes.", "For each variable, annotate the RF rank, XGB rank, and whether it belongs to the final logistic regression, CURB-65, NEWS, and the comparator XGBoost model as required, then output sorted alphabetically by variable name."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["Europe PMC search results and the target open-access article yield the candidate Wuhan COVID mortality-risk review", "article tables and discussion sections yield RF/XGBoost rankings, final logistic-regression inclusion, and comparator-variable memberships for CURB-65 / NEWS / external XGBoost", "final rows are produced only after overlapping variables and comparator-only variables are merged into one normalized variable table"], "control_dependency": ["comparator-only variables must be captured from the discussion definitions rather than from Table 2 + Table 3 alone", "comparator membership must be read from the article's explicit scheme definitions rather than inferred from common clinical usage", "variable normalization must wait until RF/XGBoost overlap and comparator-only additions are both complete"], "freeze": {"historical_window": "the 2020 open-access COVID-19 mortality-risk review about Wuhan inpatients; variable inclusion and comparator definitions are taken from the article's current full-text tables and discussion"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["First locate the target paper: a 2020 open-access review whose title contains all three terms \"prognosis model\", \"mortality risk\", and \"COVID-19\", and whose main text discusses Wuhan inpatients.", "Retain only two categories of variables: admission variables that appear in both the Random Forest and XGBoost rankings; or admission variables that, although not appearing in both rankings, belong to the CURB-65 or NEWS comparator schemes.", "For each variable, annotate the RF rank, XGB rank, and whether it belongs to the final logistic regression, CURB-65, NEWS, and the comparator XGBoost model as required, then output sorted alphabetically by variable name."], "exclusion_conditions": ["Exclude admission variables that appear in neither both RF and XGB rankings nor the CURB-65 / NEWS additional variables.", "Exclude results that stop at Table 2 + Table 3 without proceeding to the discussion to read the comparator definitions.", "Exclude results that infer comparator variables from common knowledge rather than verifying against the paper's explicit definitions."], "normalization": {"field_separator": "~", "record_separator": "|", "schema": ["variable", "RF_rank_or_0", "XGB_rank_or_0", "F", "C", "N", "X"], "variable": "use the variable wording normalized in the article tables; keep `Systolic pressure` and `Diastolic pressure` as separate variables", "rank_fields": "RF_rank_or_0 and XGB_rank_or_0 are positive integer ranks when present, otherwise 0", "boolean_fields": "F, C, N, and X are emitted as 1 or 0 only", "sorting_or_selection": "variable name alphabetical ascending", "stop_condition": ["the shared variable set from Table 2 and Table 3 is fully resolved", "the comparator definitions for CURB-65, NEWS, and the comparison XGBoost model are fixed before the final line is emitted"]}}}
{"task_id": "europemc_002-g", "domain": "EUROPEPMC_PMC", "autonomy_type": "ordered table", "oracle_output_cardinality": 6, "instruction": "Track Phase 1 studies from the 2021 Alzheimer's disease drug development pipeline review by ClinicalTrials.gov identifier, and check whether they persist in the 2022 and 2023 open-access reviews of the same series. Only include studies that were Phase 1 in the 2021 review and can still be found under the same ClinicalTrials.gov identifier in both the 2022 and 2023 reviews. Do not rely on drug names alone, as they may change. Output all qualifying studies as a single line in the format: NCT~2021 drug~2022 phase~2022 drug~2023 phase~2023 drug, sorted by NCT ascending and joined by |. If the three papers cannot be found, output NONE.", "start_url": "https://europepmc.org/advancesearch", "output_format": "Output all qualifying studies as a single line in the format: NCT~2021 drug~2022 phase~2022 drug~2023 phase~2023 drug, sorted by NCT ascending and joined by |. If no qualifying studies exist, output NONE.", "oracle_answer": "NCT03277573~Salsalate~P1~Salsalate~P1~Salsalate|NCT03634007~AAVrh.10hAPOE2~P1~LX1001~P2~LX1001|NCT03748303~Allopregnanolone (Allo)~P1~Allopregnanolone~P1~Allopregnanolone|NCT04149860~Lu AF87908~P1~Lu AF87908~P1~Lu AF87908|NCT04451408~LY3372993~P1~LY3372993~P1~LY3372993|NCT04500847~Emtricitabine~P1~Emtricitabine~P1~Emtricitabine", "metadata": {"State-Gated Retrieval": ["Only use Phase 1 studies from the 2021 Alzheimer's disease drug development pipeline article as the baseline.", "Whether a study is retained in 2022 and 2023 must be determined by exact matching on the same ClinicalTrials.gov NCT identifier, not by drug name alone.", "Only retain studies that can be found under the same NCT in both 2022 and 2023, and output them sorted by NCT ascending."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the 2021, 2022, and 2023 pipeline reviews yield the 2021 Phase 1 baseline cohort and the later-year lookup tables", "article tables yield exact ClinicalTrials.gov NCT IDs, yearly phase labels, and yearly drug names for the same study", "final rows are produced only after 2021 baseline studies are re-identified in both later reviews by exact NCT matching"], "control_dependency": ["continuation must be matched by exact NCT rather than by drug name, because drug names drift across years", "2022 and 2023 lookups must use the fixed 2021 Phase 1 NCT set as the starting cohort rather than later-year table membership alone", "yearly phase/name fields are only valid after the same NCT is confirmed across all three reviews"], "freeze": {"historical_window": "the 2021, 2022, and 2023 open-access Alzheimer's disease drug development pipeline reviews; the baseline is the 2021 Phase 1 cohort and continuation is matched by exact ClinicalTrials.gov NCT"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Only use Phase 1 studies from the 2021 Alzheimer's disease drug development pipeline article as the baseline.", "Whether a study is retained in 2022 and 2023 must be determined by exact matching on the same ClinicalTrials.gov NCT identifier, not by drug name alone.", "Only retain studies that can be found under the same NCT in both 2022 and 2023, and output them sorted by NCT ascending."], "exclusion_conditions": ["Exclude results that are continued only by drug name without exact matching on the same ClinicalTrials.gov identifier.", "Exclude studies found only in 2022 or only in 2023, not continuously retained in both years.", "Exclude studies not in the 2021 baseline Phase 1 set."], "normalization": {"field_separator": "~", "record_separator": "|", "schema": ["NCT", "agent_2021", "phase_2022", "agent_2022", "phase_2023", "agent_2023"], "NCT": "use the exact ClinicalTrials.gov identifier as the cross-year identity key", "phase_fields": "normalize phases to P1, P2, or P3; if the same NCT appears multiple times in a year, keep the highest phase for that year", "agent_fields": "keep the agent text as shown in each year-specific table", "sorting_or_selection": "NCT ascending"}}}
{"task_id": "europemc_003-g", "domain": "EUROPEPMC_PMC", "autonomy_type": "ordered table", "oracle_output_cardinality": 8, "instruction": "Identify the candidate molecules predicted in the 2023 \"Antibodies to Watch\" article to file their first marketing applications in 2022–2023, and determine which of them have actually received first approval or entered first regulatory review in the 2024 installment of the same series.\nSpecifically, the scope is limited to the open-access full texts of the 2023 and 2024 \"Antibodies to Watch\" articles in Europe PMC. Retain only those molecules that were listed in the 2023 article as **\"predicted to file first marketing applications in 2022–2023\"** and that, by the 2024 article, have either been approved or have entered first regulatory review.\nFor each qualifying molecule, output a record in the format:\nINN~2023_category~2023_prediction_label~2024_status~2024_indication~2024_region\nwhere:\n\t•\t2023_category is NC (non-cancer) or CA (cancer);\n\t•\t2024_status is A (first approval) or R (first regulatory review).\nFinally, sort records alphabetically by INN, join them with | into a single line, and output that line. If neither article can be found, output NONE.", "start_url": "https://europepmc.org/advancesearch", "output_format": "For each qualifying molecule, output: INN~2023_category~2023_prediction_label~2024_status~2024_indication~2024_region\nwhere:\n\t•\t2023_category is NC (non-cancer) or CA (cancer);\n\t•\t2024_status is A (first approval) or R (first regulatory review).\nSort records alphabetically by INN, join them with | into a single line, and output that line. If neither article can be found, output NONE.", "oracle_answer": "Cosibelimab~CA~MAA,Q2 2023~R~Squamous cell carcinoma~US|Elranatamab~CA~2023~A~Multiple myeloma~US,EU,Switzerland,Brazil|Garadacimab~NC~BLA,2023~R~Hereditary angioedema~EU|Odronextamab~CA~BLA,H2 2022~R~Diffuse large B-cell lymphoma~US,EU|Pozelimab~NC~BLA,H2 2022~A~CHAPLE disease~US|Suciraslimab~NC~NDA,2023~R~Rheumatoid arthritis~China|Talquetamab~CA~2023–25~A~Multiple myeloma~US,EU,UK,Switzerland|Zolbetuximab~CA~BLA,2023~R~HER2-negative gastric or gastroesophageal junction adenocarcinoma~US,EU,Japan,China", "metadata": {"State-Gated Retrieval": ["First, in the 2023 Antibodies to Watch article, identify all candidate molecules predicted to file first marketing applications in 2022–2023.", "Then, in the 2024 article of the same series, check each of these molecules and retain only those that have received first approval or entered first regulatory review.", "The 2024 status must distinguish A = first approval from R = first regulatory review, and the 2024 indication and region must be included."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the 2023 Antibodies to Watch article yields the forecast baseline of candidates expected to file in 2022-2023", "the 2024 article tables yield first-approval outcomes, first-regulatory-review outcomes, indications, and regions for the same INNs", "final rows are produced only after Table 1 approvals and Table 2 first-review candidates are merged back onto the 2023 forecast cohort"], "control_dependency": ["both 2024 Table 1 and Table 2 must be read so that first-review-but-not-yet-approved candidates are retained", "2024 status coding must distinguish approval from first regulatory review instead of collapsing both into a single positive match", "region and indication fields are valid only after the 2023 forecast cohort is reconciled against both 2024 tables"], "freeze": {"historical_window": "the 2023 and 2024 open-access Antibodies to Watch articles; baseline candidates come from the 2023 forecast set and 2024 outcomes come from current Table 1 plus Table 2"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["First, in the 2023 Antibodies to Watch article, identify all candidate molecules predicted to file first marketing applications in 2022–2023.", "Then, in the 2024 article of the same series, check each of these molecules and retain only those that have received first approval or entered first regulatory review.", "The 2024 status must distinguish A = first approval from R = first regulatory review, and the 2024 indication and region must be included."], "exclusion_conditions": ["Exclude results that only check approvals in 2024 Table 1 without also consulting Table 2 for first-regulatory-review candidates.", "Exclude molecules not in the 2023 forecast baseline.", "Exclude candidates that in 2024 have neither received first approval nor entered first regulatory review."], "normalization": {"field_separator": "~", "record_separator": "|", "schema": ["INN", "category_2023", "prediction_label_2023", "status_2024", "indication_2024", "region_2024"], "category_2023": "emit `NC` or `CA` exactly", "prediction_label_2023": "keep the label content inside the parentheses and remove spaces after commas, for example `BLA,H2 2022`", "status_2024": "emit `A` for first approval and `R` for first regulatory review", "region_2024": "preserve the raw region string but remove spaces after commas, for example `US,EU`", "sorting_or_selection": "INN alphabetical ascending"}}}
{"task_id": "europemc_004-g", "domain": "EUROPEPMC_PMC", "autonomy_type": "ordered table", "oracle_output_cardinality": 9, "instruction": "I am evaluating the accuracy of annual antibody watchlist predictions. Based on the Europe PMC database, I need to identify a set of candidate antibodies (INNs) whose progress has clearly lagged behind, and extract their status changes over the three-year period.\nThese target antibodies are special because their dynamics are continuously recorded across three open-access full-text articles in the Antibodies to Watch series from 2022, 2023, and 2024. Specifically, they were originally listed as \"expected to be submitted for review soon\" in the 2022 article, but due to delays, they remained as candidates on the watchlist in the 2023 article. Screening should be done at the antibody-candidate level, not by exact name string matching. If the 2022 article uses a drug code, project name, or other identifiable name, and the 2023 or 2024 article explicitly provides its correspondence to a specific INN, it should be treated as the same candidate entity, and the confirmed INN should be used in the final output.\nFor these continuously delayed candidate antibodies, I need to determine their final status in the 2024 article: whether they finally received first approval, are in first regulatory review, remain on a subsequent watchlist, or have completely disappeared from the main 2024 article list.\nPlease help me identify this set of antibodies and output the extracted parameters as a single-line string joined by |, strictly sorted alphabetically by INN. If the three target articles cannot all be found in the database, output NONE. The internal field order for each antibody must be fixed as: INN~2022 category~2022 phase~2023 prediction label~2024 status~2024 phase or NA~2024 indication or NA~2024 region or NA. No extra spaces should be added around the delimiters in the output, but original spaces within fields should be preserved. Use CA/NC uniformly for 2022 category, P2/P3 for phase, and A/R/F/D for 2024 status, representing First approved, In first regulatory review, subsequent watchlist, and disappeared from the 2024 main article list, respectively.", "start_url": "https://europepmc.org/advancesearch", "output_format": "Output the extracted parameters as a single-line string joined by |, strictly sorted alphabetically by INN. If the three target articles cannot all be found in the database, output NONE. The internal field order for each antibody must be fixed as: INN~2022 category~2022 phase~2023 prediction label~2024 status~2024 phase or NA~2024 indication or NA~2024 region or NA. No extra spaces should be added around the delimiters in the output, but original spaces within fields should be preserved. Use CA/NC uniformly for 2022 category, P2/P3 for phase, and A/R/F/D for 2024 status, representing First approved, In first regulatory review, subsequent watchlist, and disappeared from the 2024 main article list, respectively.", "oracle_answer": "Apamistamab-I-131~CA~P3~BLA,H1 2023~F~P3~Ablation of bone marrow prior to transplantation in AML patients~NA|Bentracimab~NC~P3~BLA,Q4 2022~F~P3~Reversal of the antiplatelet effects of ticagrelor~NA|Cosibelimab~CA~P3~MAA,Q2 2023~R~NA~Squamous cell carcinoma~US|Erfonrilimab~CA~P3~NDA~F~P3~Pancreatic ductal adenocarcinoma~NA|Odronextamab~CA~P2~BLA,H2 2022~R~NA~Diffuse large B-cell lymphoma~US,EU|Talquetamab~CA~P2~2023-25~A~NA~Multiple myeloma~US,EU,UK,Switzerland|Tiragolumab~CA~P3~2023~F~P3~NSCLC~NA|Zanidatamab~CA~P2~2023~D~NA~NA~NA|Zolbetuximab~CA~P3~BLA,2023~R~NA~HER2-negative gastric or gastroesophageal junction adenocarcinoma~US,EU,Japan,China", "metadata": {"State-Gated Retrieval": ["Start with the 2022 Antibodies to Watch candidates listed as \"expected to be submitted for review soon\" as the baseline, but retain only those molecules that still remain as watchlist candidates in the 2023 same-series article.", "For these continuously delayed candidates, then determine their final status in the 2024 article: first approved, in first regulatory review, or completely absent from Tables 1-4.", "Name comparison must handle the Apamistamab naming drift; complete absence in 2024 must be explicitly encoded as D."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the 2022, 2023, and 2024 Antibodies to Watch articles yield the 2022 soon-to-file baseline, 2023 carry-over membership, and 2024 status tables", "article tables and full-text naming context yield name normalizations, 2024 table presence/absence, and final status labels for each carried-over INN", "final rows are produced only after 2022 candidates are filtered by 2023 carry-over status and then resolved against 2024 Tables 1-4 plus explicit absence"], "control_dependency": ["2022 candidates must first be filtered by 2023 carry-over status before they are matched against the 2024 tables", "Apamistamab naming drift forces explicit normalization before 2024 lookup", "complete 2024 absence must be emitted as D rather than treated as a lookup failure"], "freeze": {"historical_window": "the 2022, 2023, and 2024 open-access Antibodies to Watch articles; the 2022 baseline is filtered by 2023 carry-over status and then checked against 2024 Tables 1-4, with explicit absence retained as a status"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Start with the 2022 Antibodies to Watch candidates listed as \"expected to be submitted for review soon\" as the baseline, but retain only those molecules that still remain as watchlist candidates in the 2023 same-series article.", "For these continuously delayed candidates, then determine their final status in the 2024 article: first approved, in first regulatory review, or completely absent from Tables 1-4.", "Name comparison must handle the Apamistamab naming drift; complete absence in 2024 must be explicitly encoded as D."], "exclusion_conditions": ["Exclude results that directly match the 2022 baseline against 2024 by literal string matching without first applying the 2023 carry-over filter.", "Exclude matching results that do not handle the Apamistamab naming drift.", "Exclude cases where complete absence in 2024 is treated as a lookup failure rather than a status that must be output."], "normalization": {"field_separator": "~", "record_separator": "|", "schema": ["INN", "category_2022", "phase_2022", "prediction_label_2023", "status_2024", "phase_or_NA_2024", "indication_or_NA_2024", "region_or_NA_2024"], "category_2022": "emit `NC` or `CA` exactly", "phase_2022": "normalize stage labels such as `Phase 3` to `P3` and `Pivotal Phase 2` to `P2`", "prediction_label_2023": "keep the label content inside the parentheses and remove spaces after commas, for example `BLA,H1 2023`", "status_2024": "emit one of `A`, `R`, `F`, or `D` exactly", "conditional_fields_2024": "if status is A or R, phase_or_NA_2024 must be `NA` and region_or_NA_2024 must carry the raw region string; if status is F, region_or_NA_2024 is `NA` and phase_or_NA_2024 carries the normalized phase; if status is D, all three trailing 2024 fields are `NA`", "naming_exception": "preserve the fixed source exception around the Apamistamab-I-131 naming variant instead of over-normalizing the INN", "sorting_or_selection": "INN alphabetical ascending"}}}
{"task_id": "europemc_005-g", "domain": "EUROPEPMC_PMC", "autonomy_type": "ordered table", "oracle_output_cardinality": 20, "instruction": "I am compiling a cross-platform summary of predictors for COVID-19 inpatient outcome trajectories. This task centers on a 2024 WHO global individual patient data validation paper and its three linked original model papers. In this 2024 validation paper, four prognostic models ultimately completed external validation in low- and middle-income country data. We need to identify which distinct admission variables each of these four models used in their actual evaluations.\nFor each admission variable identified, we need to determine its characteristics in actual use: first, its matching difficulty in the WHO platform baseline table, i.e., whether it can be matched directly under the same or a substantively equivalent name or must go through a fixed mapping; second, its source background, i.e., whether it belongs to the model with the strongest performance in validation, the model with calibration closest to ideal, or an original model that explicitly used dynamic data.\nTo ensure consistency, model abbreviations, performance attributions, dynamic-data attributes, and variable conflicts shall be based on verifiable model descriptions, summary performance results, method explanations, and field definitions in the target literature; any variable naming or mapping differences shall be marked with a conservative correspondence that maintains cross-model interpretability.\nIf all four articles (the one validation paper and the three original model papers) can be found in the database, please organize all extracted variable information and output it as a single string with no spaces, joining records by | in strict alphabetical order of variable names (if the four articles cannot be found, output NONE).\nFormat requirement: each variable's internal information must be written in the fixed format variable~category~B~W~Zh~Zo~WHOfield~WHOx~TopAUC~BestCal~Dyn~Conflict. Boolean states use 1 for yes and 0 for no; WHO baseline table matching uses E for a direct or substantively equivalent-name match and M for fixed mapping required.", "start_url": "https://europepmc.org/advancesearch", "output_format": "If all four articles (the one validation paper and the three original model papers) can be found in the database, organize all extracted variable information and output it as a single string with no spaces, joining records by | in strict alphabetical order of variable names; if the four articles cannot be found, output NONE. Each variable's internal information must be written in the fixed format variable~category~B~W~Zh~Zo~WHOfield~WHOx~TopAUC~BestCal~Dyn~Conflict. Boolean states use 1 for yes and 0 for no; WHO baseline table matching uses E for a direct or substantively equivalent-name match and M for fixed mapping required.", "oracle_answer": "Age~DEM~1~1~1~1~Age~E~1~1~1~0|Cancer~COM~0~0~1~0~Cancer~E~1~0~0~0|Chronic kidney disease~COM~0~0~0~1~CKD~M~0~0~0~1|Chronic lung disease~COM~0~0~1~0~Pulmonary disease~M~1~0~0~0|Chronic renal disease~COM~0~0~1~0~CKD~M~1~0~0~1|Coronary heart disease~COM~0~1~0~0~CHD~M~0~0~0~1|Cough~SYM~0~0~1~0~Cough~E~1~0~0~0|Diabetes mellitus~COM~0~0~1~0~Diabetes~M~1~0~0~0|Diarrhea~SYM~0~0~1~0~Diarrhea~E~1~0~0~0|Dyspnea~SYM~0~0~1~0~Dyspnea~E~1~0~0~0|Fever~VIT~0~0~0~1~Body temperature~M~0~0~0~0|Heart disease~COM~0~0~1~0~CHD~M~1~0~0~1|Hypertension~COM~0~1~1~0~Hypertension~E~1~0~0~0|Immunocompromised status~COM~0~0~1~0~Immunosuppression~M~1~0~0~0|O2 saturation~VIT~1~0~0~0~Oxygen saturation~M~0~1~1~0|Respiratory rate~VIT~1~0~0~1~Respiratory rate~E~0~1~1~0|Sex~DEM~0~0~1~0~Men~M~1~0~0~0|Smoking~DEM~1~0~0~0~Smoking~E~0~1~1~1|Smoking status~DEM~0~0~0~1~Smoking~M~0~0~0~1|Systolic blood pressure~VIT~0~0~0~1~Systolic blood pressure~E~0~0~0~0", "metadata": {"State-Gated Retrieval": ["First, in the 2024 WHO global individual patient data validation paper, identify the four prognostic models that successfully completed external validation in LMIC data, and trace back to the three corresponding original model papers.", "For each predictor, retain the original wording from WHO Table 1, and annotate its matching method in the WHO baseline table as E or M, and whether it comes from the TopAUC, BestCal, or Dyn model.", "When encountering a predictor that is conceptually merged on the WHO side, trace back to the original Wang/Zhou papers to split it apart before recalculating the relevant fields."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the 2024 WHO global validation paper yields the four externally validated models and the WHO-side predictor wording/mapping baseline", "the three original model papers yield model-specific predictor wordings, dynamic-data flags, and conflicts hidden by WHO-side conceptual merging", "final rows are produced only after WHO predictors are kept separate, traced back to the original papers, and then re-annotated for mapping and model provenance"], "control_dependency": ["conceptually merged WHO predictors must be split back out when the original Wang/Zhou papers expose distinct variables", "Table 2 mapping decisions are not valid until Table 1 wording and original-paper predictor sets are reconciled", "TopAUC / BestCal / Dyn flags must be determined after the predictor split is applied correctly"], "freeze": {"historical_window": "the 2024 WHO global IPD validation paper and the three linked original COVID prognosis-model papers, using their current open-access full texts and tables"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["First, in the 2024 WHO global individual patient data validation paper, identify the four prognostic models that successfully completed external validation in LMIC data, and trace back to the three corresponding original model papers.", "For each predictor, retain the original wording from WHO Table 1, and annotate its matching method in the WHO baseline table as E or M, and whether it comes from the TopAUC, BestCal, or Dyn model.", "When encountering a predictor that is conceptually merged on the WHO side, trace back to the original Wang/Zhou papers to split it apart before recalculating the relevant fields."], "exclusion_conditions": ["Exclude results that directly merge conceptually similar predictors after WHO Table 1 without tracing back to the original papers for splitting.", "Exclude results that rely solely on WHO Table 2 for loose mapping without incorporating the original paper wording.", "Exclude results that fix TopAUC / BestCal / Dyn annotations before the predictor split correction is applied."], "normalization": {"field_separator": "~", "record_separator": "|", "schema": ["variable", "category", "B", "W", "Zh", "Zo", "WHOfield", "WHOx", "TopAUC", "BestCal", "Dyn", "Conflict"], "variable_set": "the candidate variable set comes only from the WHO validation Table 1 and remains at the original 20 predictors", "conflict_groups": "do not pre-merge the three source conflict groups `Smoking`/`Smoking status`, `Chronic renal disease`/`Chronic kidney disease`, and `Coronary heart disease`/`Heart disease` before applying the conflict flag", "boolean_fields": "B, W, Zh, Zo, TopAUC, BestCal, Dyn, and Conflict are emitted as 1 or 0 only", "WHOx": "emit `E` for exact WHO matches or direct matches under a substantively equivalent name, `M` for the fixed mapped match, and `NA` when no WHO field matches", "sorting_or_selection": "variable name alphabetical ascending"}}}
{"task_id": "genome_001-g", "domain": "KEGG_BRITE", "autonomy_type": "ordered table", "oracle_output_cardinality": 3, "instruction": "From KEGG, identify immunosuppressants used for ulcerative colitis that have been approved in Japan by the end of June 2025 and carry clear cytochrome P450 (CYP) metabolism or inhibition evidence. For each qualifying drug, retain its Japanese therapeutic classification code, ATC code, corresponding target branch, and Japan first approval date. Output a single-line string sorted by Japan first approval date from earliest to latest, formatted as <DRUG_ID>|<JP_class>|<ATC>|<Target_branch>|<JP_date>; separate multiple entries with commas; if no qualifying items exist, output NONE.", "start_url": "https://www.genome.jp/kegg/brite.html", "output_format": "Output a single-line string sorted by Japan first approval date from earliest to latest, formatted as <DRUG_ID>|<JP_class>|<ATC>|<Target_branch>|<JP_date>; separate multiple entries with commas; if no qualifying items exist, output NONE.", "oracle_answer": "D09783|3999|L04AF01|JAK1/JAK2/JAK3/TYK2|2013/3/25,D11048|3999|L04AF03|JAK1|2020/1/23,D10931|2399|L04AE05|S1PR1/S1PR4/S1PR5|2025/6/24", "metadata": {"State-Gated Retrieval": ["Within the scope of KEGG immunosuppressants for ulcerative colitis, retain only those approved in Japan by 2025-06-30 and carrying clear CYP-related metabolism information or CYP inhibition evidence.", "Each candidate drug must be re-verified against the corrected UC drug list for Japanese therapeutic classification, ATC, target branch, CYP evidence, and Japan first approval date.", "Finally, output sorted by Japan first approval date from earliest to latest."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["KEGG disease H01466 and linked drug entries yield the ulcerative-colitis immunosuppressant candidate set", "each candidate drug entry yields Japanese class, ATC, target branch, CYP metabolism/inhibition evidence, and Japan approval date", "final rows are produced only after the full UC candidate set is determined and every candidate is checked against CYP evidence and PMDA timing"], "control_dependency": ["the candidate set must include Japanese class 3999 UC drugs when the disease page exposes them rather than being limited to 2399 alone", "target / CYP / PMDA validation must be applied across the full candidate set rather than only the old 2399 subset", "entries without explicit CYP evidence may be dropped only after the full candidate set has been established"], "freeze": {"historical_window": "KEGG ulcerative-colitis drug entries as objectively existing drug entities, using target, CYP, Japan-class, and related entry metadata as relatively stable annotations"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Within the scope of KEGG immunosuppressants for ulcerative colitis, retain only those approved in Japan by 2025-06-30 and carrying clear CYP-related metabolism information or CYP inhibition evidence.", "Each candidate drug must be re-verified against the corrected UC drug list for Japanese therapeutic classification, ATC, target branch, CYP evidence, and Japan first approval date.", "Finally, output sorted by Japan first approval date from earliest to latest."], "exclusion_conditions": ["Exclude results that incorrectly lock the upstream scope to the 2399 digestive branch without falling back to the full UC drug list.", "Exclude entries without clear CYP-related metabolism information or CYP inhibition evidence.", "Exclude entries with a Japan approval date after 2025-06-30 or that are not within the KEGG UC immunosuppressant scope."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["DRUG_ID", "JP_class", "ATC", "Target_branch", "JP_date"], "dedup_key": "DRUG_ID", "JP_date": "use the canonical KEGG-style date format YYYY/M/D with no zero-padding requirement beyond the source value", "Target_branch": "emit the most specific target branch visible in the source path; when multiple parallel branches survive, join them with `/`", "sorting_or_selection": {"primary": "JP_date ascending", "secondary": "DRUG_ID ascending"}}}}
{"task_id": "genome_002-g", "domain": "KEGG_GENOME", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "From the KEGG host–symbiont examples, identify animal endosymbiosis cases whose bacterial symbiont KEGG GENOME entry was already recorded by the end of 2005 and whose bacterial symbiont genome contains a clearly named extra replicon. Additionally, the names of these replicons must directly correspond to amino acid biosynthesis modules on the host–symbiont joint pathway map. For the surviving case(s), list all qualifying pathway–module–replicon combinations. Output a single-line string in the format <group>|<symbiont>|<pathway>|<module>|<replicon>|<created_year>, sorted by pathway ID in ascending order, with entries separated by commas. If no qualifying items exist, output NONE. Use KEGG organism codes for species identifiers.", "start_url": "https://www.genome.jp/kegg/genome/", "output_format": "Output a single-line string in the format <group>|<symbiont>|<pathway>|<module>|<replicon>|<created_year>, sorted by pathway ID in ascending order, with entries separated by commas. If no qualifying items exist, output NONE. Use KEGG organism codes for species identifiers.", "oracle_answer": "api+buc|buc|00290|M00432|pLeu|2000,api+buc|buc|00400|M00023|pTrp|2000", "metadata": {"State-Gated Retrieval": ["Retain only host–symbiont examples where the bacterial symbiont KEGG GENOME Created year is no later than 2005 and the genome carries a named extra replicon.", "The names of these replicons must directly correspond to amino acid biosynthesis modules on the host–symbiont joint pathway page.", "Finally, output all qualifying pathway–module–replicon records from the surviving case(s), sorted by pathway ID in ascending order."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["KEGG host–symbiont examples and symbiont genome pages yield the pre-2006 candidate pairs and named extra replicons", "Joint pathway maps, module pages, and genome annotations yield the pathway–module–replicon links attributable to the symbiont side", "Final rows are produced only after the surviving host–symbiont pair is validated against both replicon existence and module support on the joint pathway page"], "control_dependency": ["Only branches satisfying the named-plasmid requirement may be retained, so the bmy+wbm branch must be excluded when the wbm genome page fails that requirement", "Module retrieval must use the api+buc pathway state and the 00400 view, because that view is the one that actually exposes M00023", "Module support is valid only after symbiont-side evidence for both M00023 and M00432 is validated in the selected pathway state"], "freeze": {"historical_window": "KEGG host–symbiont examples as objectively existing biological entities, using joint pathway, module, and genome replicon annotations as relatively stable structural information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Retain only host–symbiont examples where the bacterial symbiont KEGG GENOME Created year is no later than 2005 and the genome carries a named extra replicon.", "The names of these replicons must directly correspond to amino acid biosynthesis modules on the host–symbiont joint pathway page.", "Finally, output all qualifying pathway–module–replicon records from the surviving case(s), sorted by pathway ID in ascending order."], "exclusion_conditions": ["Exclude cases like bmy+wbm where the symbiont genome page does not satisfy the named-plasmid requirement.", "Exclude results that only follow pathway names along pages such as 00380 without returning to the joint pathway page to verify module evidence.", "Exclude records where the replicon–module relationship cannot be directly established on the joint pathway map."], "normalization": {"field_separator": "|", "record_separator": ",", "schema": ["group", "symbiont", "pathway", "module", "replicon", "created_year"], "group": "Use the canonical host–symbiont group label from the source workflow, such as `api+buc`", "symbiont": "Use the bare lowercase KEGG organism code", "pathway": "Emit the numeric KEGG pathway ID without a `map` prefix", "module": "Normalize the KEGG module identifier to the `Mxxxxx` form", "replicon": "Preserve the named extra replicon exactly as displayed in the source, such as `pLeu` or `pTrp`", "created_year": "Emit the four-digit created year", "sorting_or_selection": "Pathway numeric ascending"}}}
{"task_id": "genome_003-g", "domain": "KEGG_GENOME", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I am preparing a handout for a primate evolution course and want a two-row comparison table for this vitamin C metabolism exception pattern. Output two rows with columns rank|clade_role|parent_clade|clade|anomaly_pathway|contrast_pathway|retained_upstream_module|vitamin_c_synthesis_module|vitamin_c_module_status|example_org|cutoff_year. The first row should be EXCEPTION_CLADE and the second row RETAINED_SISTER_CLADE. Use KEGG organism codes for the example species.", "start_url": "https://www.genome.jp/kegg/kegg2.html", "output_format": "Output a two-row comparison table with columns rank|clade_role|parent_clade|clade|anomaly_pathway|contrast_pathway|retained_upstream_module|vitamin_c_synthesis_module|vitamin_c_module_status|example_org|cutoff_year.", "oracle_answer": "1|EXCEPTION_CLADE|Primates|Haplorrhini|map00053|map00040|M00014|M00129|LOST|csyr|2022\n2|RETAINED_SISTER_CLADE|Primates|Strepsirrhini|map00053|map00040|M00014|M00129|RETAINED|oga|2022", "metadata": {"State-Gated Retrieval": ["Search only among primate genome entries established by the end of 2022, and isolate the anomalous lower-level clade that retains the upstream glucuronate module M00014 but lacks the animal-type vitamin C biosynthesis module M00129.", "That anomalous clade must be paired with a sister branch under the same higher-level split, and the sister branch must still retain M00129.", "The final output must place the anomalous branch and the retained sister branch into a two-row comparison table, including the required pathway, module, and example-species fields."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the KEGG primate clade pages and genome entries yield candidate subclades and example species established through the cutoff year", "the pathway and module views yield M00014 and M00129 presence/absence evidence for the exception clade and its sister branch", "final rows are produced only after the exception clade and retained sister branch are jointly validated under the same parent split and paired with example organisms"], "control_dependency": ["the exception clade remains provisional until both the sister-branch contrast and the example-species checks succeed", "the comparison view must make retained M00014 and missing M00129 visible together", "the two output rows are emitted together only after the exception clade and retained sister branch are fixed under the same parent split"], "freeze": {"historical_window": "the objectively existing primate genome entries established on or before 2022, using current KEGG clade, pathway, and module pages as relatively stable reference information for comparing retained M00014 against lost M00129"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Search only among primate genome entries established by the end of 2022, and isolate the anomalous lower-level clade that retains the upstream glucuronate module M00014 but lacks the animal-type vitamin C biosynthesis module M00129.", "That anomalous clade must have a contrastable sister branch, and that sister branch must still retain M00129.", "The final output must place the anomalous branch and the retained sister branch into a two-row comparison table, including the required pathway, module, and example-species fields."], "exclusion_conditions": ["Exclude results that stop at Simiiformes without checking the broader boundary and therefore miss Haplorrhini.", "Exclude pathway views that do not make the M00014 versus M00129 contrast visible together.", "Exclude candidate clades that have not been validated against a retained sister branch."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["rank", "clade_role", "parent_clade", "clade", "anomaly_pathway", "contrast_pathway", "retained_upstream_module", "vitamin_c_synthesis_module", "vitamin_c_module_status", "example_org", "cutoff_year"], "rank": "emit 1 for EXCEPTION_CLADE and 2 for RETAINED_SISTER_CLADE", "clade_role": ["EXCEPTION_CLADE", "RETAINED_SISTER_CLADE"], "pathway_normalization": "emit KEGG pathway ids with the map prefix", "module_normalization": "emit KEGG module ids with the leading M", "vitamin_c_module_status": ["LOST", "RETAINED"], "organism_code": "use bare KEGG organism codes in lowercase, not species names", "cutoff_year": "use a four-digit year; this task is frozen at 2022", "sorting_or_selection": "emit rows in the fixed order EXCEPTION_CLADE then RETAINED_SISTER_CLADE"}}}
{"task_id": "genome_004-g", "domain": "KEGG_BRITE", "autonomy_type": "ordered table", "oracle_output_cardinality": 8, "instruction": "I am compiling an ordered table of systemic drugs approved for plaque psoriasis in Japan. Include only active ingredients that were approved in Japan by the end of 2024, are used for systemic treatment of plaque psoriasis, and target the IL-17/IL-23 signaling axis. Exclude topical drugs and systemic drugs acting through other axes. Output multiple rows sorted by Japan first approval date from earliest to latest; for drugs approved on the same day, sort by DRUG_ID ascending. Each row format: <DRUG_ID>|<Name>|<ATC>|<Target_branch>|<PMDA_date>.", "start_url": "https://www.kegg.jp/brite/br08301", "output_format": "Output multiple rows sorted by Japan first approval date from earliest to latest; for drugs approved on the same day, sort by DRUG_ID ascending. Each row format: <DRUG_ID>|<Name>|<ATC>|<Target_branch>|<PMDA_date>.", "oracle_answer": "D09967|Secukinumab|L04AC10|IL17A|2014/12/26\nD10061|Brodalumab|L04AC12|IL17RA|2016/7/4\nD10071|Ixekizumab|L04AC13|IL17A|2016/7/4\nD10438|Guselkumab|L04AC16|IL23A|2018/3/23\nD11052|Risankizumab|L04AC18|IL23A|2019/3/26\nD10400|Tildrakizumab|L04AC17|IL23A|2020/6/29\nD11550|Bimekizumab|L04AC21|IL17A/IL17F|2022/1/20\nD11817|Deucravacitinib|L04AF07|TYK2|2022/9/26", "metadata": {"State-Gated Retrieval": ["Retain only active ingredients that were approved in Japan by 2024-12-31, are used for systemic treatment of plaque psoriasis, and genuinely act on the IL-17/IL-23 signaling axis.", "The candidate set must cover the ligand side, receptor side, and TYK2 kinase side, not only the interleukin ligand side.", "Final entries sorted by Japan first approval date ascending; for the same date, sort by DRUG_ID ascending."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["KEGG disease/drug entry points yield the Japan-approved systemic plaque-psoriasis drug cohort through 2024-12-31", "each drug entry yields target branch, ATC, systemic-use validity, and Japan approval date for IL-17/IL-23-axis candidates", "final rows are produced only after ligand-side, receptor-side, and kinase-side candidates are jointly reconciled and re-sorted by Japan approval date"], "control_dependency": ["entry-point coverage must include receptor-side IL17RA and kinase-side TYK2 mechanisms when downstream drug pages expose them, rather than staying ligand-only", "Ustekinumab must be removed only when target interpretation shows IL12 + IL23A rather than the IL-17/IL-23-axis criterion required here", "systemic-use, Japanese 3999, and PMDA-date filtering must be applied after each expansion of the candidate set rather than attached to the earlier ligand-only state"], "freeze": {"historical_window": "KEGG plaque-psoriasis systemic-drug entries as objectively existing drug entities, using target-branch and related drug-entry metadata as relatively stable annotations"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Retain only active ingredients that were approved in Japan by 2024-12-31, are used for systemic treatment of plaque psoriasis, and genuinely act on the IL-17/IL-23 signaling axis.", "The candidate set must cover the ligand side, receptor side, and TYK2 kinase side, not only the interleukin ligand side.", "Final entries sorted by Japan first approval date ascending; for the same date, sort by DRUG_ID ascending."], "exclusion_conditions": ["Exclude topical drugs, non-systemic drugs, or entries outside the specified historical freeze window.", "Exclude Ustekinumab, whose actual target is IL12 + IL23A, because it does not meet the IL-17/IL-23 axis criterion required here.", "Exclude results that miss Brodalumab or Deucravacitinib due to considering only the ligand side."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["DRUG_ID", "Name", "ATC", "Target_branch", "PMDA_date"], "PMDA_date": "use the canonical KEGG-style date format YYYY/M/D with no zero-padding requirement beyond the source value", "Target_branch": "emit the most specific target branch visible in the source path; when multiple parallel branches survive, join them with `/`", "sorting_or_selection": {"primary": "PMDA_date ascending", "secondary": "DRUG_ID ascending"}}}}
{"task_id": "nvd_001-g", "domain": "NVD", "autonomy_type": "ordered table", "instruction": "Identify which milestone versions of Apache Tomcat 9.0.0, released before the final GA release, are both affected by all four specified CVEs and have a CPE name notation change in NVD history. A CPE name notation change means the same milestone version was recorded as mN in older records but changed to milestoneN in current records. The search scope is limited to Apache Tomcat 9.0.0 milestone versions; these versions must appear in the current NVD affected product configurations for CVE-2016-5018, CVE-2016-6794, CVE-2016-6796, and CVE-2016-6797. Exclude milestone versions brought in only by other Tomcat 9.0.0 pre-release related vulnerabilities, and also exclude Tomcat 6, Tomcat 7, Tomcat 8, Tomcat 8.5, and all non-Tomcat products. Output a multi-row ordered table with columns: milestone | legacy_cpe | current_cpe | required_cves | remap_time_nvd, sorted by milestone number in ascending order.", "start_url": "https://nvd.nist.gov/vuln/search#/nvd/home?resultType=records", "output_format": "Output a multi-row ordered table with columns: milestone | legacy_cpe | current_cpe | required_cves | remap_time_nvd, sorted by milestone number in ascending order.", "oracle_output_cardinality": 3, "oracle_answer": "CVE:2018-7543|1.2.32|lite|Reanalysis@2021-10-18|snapcreek->awesomemotive@2026-02-02,CVE:2018-17207|1.2.42|lite|Reanalysis@2021-10-18|snapcreek->awesomemotive@2026-02-02,CVE:2020-11738|1.3.28|lite+pro|ModifiedAnalysis@2021-10-18|snapcreek->awesomemotive@2026-02-02", "metadata": {"State-Gated Retrieval": ["Retain only Duplicator-related CVEs whose current product line belongs to awesomemotive:duplicator and whose deprecated vendor chain traces back to snapcreek:duplicator.", "Each candidate CVE must satisfy all of the following in its Change History: the Initial Analysis was not yet lite/lite+pro; the first change to lite or lite+pro occurred in a NIST analysis event on 2021-10-18; and a snapcreek -> awesomemotive vendor remap occurred on 2026-02-02.", "The final output retains only CVE rows that meet all three conditions above."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["NVD product/CPE search pages yield the current awesomemotive:duplicator family and the deprecated snapcreek:duplicator lineage", "each target CVE detail page and Change History yield the Initial Analysis state, the first lite/lite+pro analysis event, and the 2026-02-02 vendor-remap event", "final rows are produced only after candidate CVEs are intersected on both snapcreek -> awesomemotive migration and later introduction of lite/lite+pro in NIST analysis history"], "control_dependency": ["the search must include the deprecated snapcreek lineage rather than only the current product state", "vendor migration alone is insufficient until detail-level history confirms that lite/lite+pro was introduced after Initial Analysis", "candidate CVEs must be filtered against the full change-history chronology, excluding any whose Initial Analysis already contained lite"], "freeze": {"historical_window": "Duplicator CVEs published from 2018-01-01 through 2024-12-31, evaluated against NVD analysis history up to the 2026-02-02 snapcreek -> awesomemotive vendor-remap events"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Retain only Duplicator-related CVEs whose current product line belongs to awesomemotive:duplicator and whose deprecated vendor chain traces back to snapcreek:duplicator.", "Each candidate CVE must satisfy all of the following in its Change History: the Initial Analysis was not yet lite/lite+pro; the first change to lite or lite+pro occurred in a NIST analysis event on 2021-10-18; and a snapcreek -> awesomemotive vendor remap occurred on 2026-02-02.", "The final output retains only CVE rows that meet all three conditions above."], "exclusion_conditions": ["Exclude results that rely only on the current description or current product state without tracing back to the deprecated snapcreek chain and Change History.", "Exclude CVEs whose Initial Analysis was already lite/lite+pro.", "Exclude records without a 2026-02-02 vendor remap or that do not belong to the Duplicator product line."], "normalization": {"field_separator": "|", "record_separator": ",", "schema_note": "the current source docx rubric, evidence, and oracle answer use a Duplicator CVE rowset, even though the instruction/output_format text still describes a Tomcat milestone task; normalization therefore follows the checked oracle row schema rather than the stale milestone wording", "schema": ["CVE", "lite_upper_version", "lite_vs_lite+pro", "analysis_event@date", "snapcreek_to_awesomemotive@date"], "CVE": "emit the identifier in the canonical `CVE:YYYY-NNNN` form used by the oracle", "lite_vs_lite+pro": "emit `lite` or `lite+pro` exactly", "analysis_event@date": "emit `Reanalysis@YYYY-MM-DD` or `ModifiedAnalysis@YYYY-MM-DD` exactly as supported by the source record", "snapcreek_to_awesomemotive@date": "emit the ownership-transfer event as `snapcreek->awesomemotive@YYYY-MM-DD`", "sorting_or_selection": {"primary": "NVD published date ascending", "secondary": "CVE ascending"}}}}
{"task_id": "nvd_002-g", "instruction": "We are compiling an exception list of legacy MySQL 5.1 assets left after Oracle's acquisition of MySQL. We need to identify which old CPE 2.3 entries whose name prefix still reads mysql:mysql were explicitly remapped by NIST to the corresponding oracle:mysql entries in the NVD history records for four InnoDB denial-of-service vulnerabilities. Start with CVE-2010-3676, CVE-2010-3677, CVE-2010-3678, and CVE-2010-3680. For each vulnerability, examine the relevant CPE change records in NVD and find MySQL 5.1 entries where the prefix changed from mysql:mysql to oracle:mysql in all four vulnerabilities. During screening, retain only records that underwent this legacy-to-current prefix migration; exclude MySQL 5.0 branch entries and entries that only performed version alias normalization within the old mysql:mysql prefix. Finally, output a multi-row ordered table with columns version_update | legacy_cpe | current_cpe | remap_time_nvd, sorted by numeric version in ascending order; within the same version, records with an empty update field come before those with a non-empty update.", "domain": "NVD", "autonomy_type": "ordered table", "oracle_output_cardinality": 44, "start_url": "https://nvd.nist.gov/products/cpe/search", "output_format": "Output a multi-row ordered table with columns version_update | legacy_cpe | current_cpe | remap_time_nvd, sorted by numeric version in ascending order; within the same version, entries without an update come before entries with an update.", "oracle_answer": "version_update|legacy_cpe|current_cpe|remap_time_nvd\n5.1.1|cpe:2.3:a:mysql:mysql:5.1.1:::::::*|cpe:2.3:a:oracle:mysql:5.1.1:::::::*|2019-12-17 15:04:33\n5.1.2|cpe:2.3:a:mysql:mysql:5.1.2:::::::*|cpe:2.3:a:oracle:mysql:5.1.2:::::::*|2019-12-17 15:04:39\n5.1.3|cpe:2.3:a:mysql:mysql:5.1.3:::::::*|cpe:2.3:a:oracle:mysql:5.1.3:::::::*|2019-12-17 15:04:46\n5.1.4|cpe:2.3:a:mysql:mysql:5.1.4:::::::*|cpe:2.3:a:oracle:mysql:5.1.4:::::::*|2019-12-17 15:04:53\n5.1.10|cpe:2.3:a:mysql:mysql:5.1.10:::::::*|cpe:2.3:a:oracle:mysql:5.1.10:::::::*|2019-12-17 15:05:30\n5.1.11|cpe:2.3:a:mysql:mysql:5.1.11:::::::*|cpe:2.3:a:oracle:mysql:5.1.11:::::::*|2019-12-17 15:05:37\n5.1.12|cpe:2.3:a:mysql:mysql:5.1.12:::::::*|cpe:2.3:a:oracle:mysql:5.1.12:::::::*|2019-12-17 15:05:45\n5.1.13|cpe:2.3:a:mysql:mysql:5.1.13:::::::*|cpe:2.3:a:oracle:mysql:5.1.13:::::::*|2019-12-17 15:05:52\n5.1.14|cpe:2.3:a:mysql:mysql:5.1.14:::::::*|cpe:2.3:a:oracle:mysql:5.1.14:::::::*|2019-12-17 15:05:59\n5.1.15|cpe:2.3:a:mysql:mysql:5.1.15:::::::*|cpe:2.3:a:oracle:mysql:5.1.15:::::::*|2019-12-17 15:06:07\n5.1.16|cpe:2.3:a:mysql:mysql:5.1.16:::::::*|cpe:2.3:a:oracle:mysql:5.1.16:::::::*|2019-12-17 15:06:14\n5.1.17|cpe:2.3:a:mysql:mysql:5.1.17:::::::*|cpe:2.3:a:oracle:mysql:5.1.17:::::::*|2019-12-17 15:06:21\n5.1.18|cpe:2.3:a:mysql:mysql:5.1.18:::::::*|cpe:2.3:a:oracle:mysql:5.1.18:::::::*|2019-12-17 15:06:27\n5.1.19|cpe:2.3:a:mysql:mysql:5.1.19:::::::*|cpe:2.3:a:oracle:mysql:5.1.19:::::::*|2019-12-17 15:06:34\n5.1.20|cpe:2.3:a:mysql:mysql:5.1.20:::::::*|cpe:2.3:a:oracle:mysql:5.1.20:::::::*|2019-12-17 15:06:42\n5.1.21|cpe:2.3:a:mysql:mysql:5.1.21:::::::*|cpe:2.3:a:oracle:mysql:5.1.21:::::::*|2019-12-17 15:06:50\n5.1.22|cpe:2.3:a:mysql:mysql:5.1.22:::::::*|cpe:2.3:a:oracle:mysql:5.1.22:::::::*|2019-12-17 15:06:56\n5.1.23:a|cpe:2.3:a:mysql:mysql:5.1.23:a::::::|cpe:2.3:a:oracle:mysql:5.1.23:a::::::|2019-12-17 15:23:28\n5.1.24|cpe:2.3:a:mysql:mysql:5.1.24:::::::*|cpe:2.3:a:oracle:mysql:5.1.24:::::::*|2019-12-17 15:07:03\n5.1.25|cpe:2.3:a:mysql:mysql:5.1.25:::::::*|cpe:2.3:a:oracle:mysql:5.1.25:::::::*|2019-12-17 15:07:10\n5.1.26|cpe:2.3:a:mysql:mysql:5.1.26:::::::*|cpe:2.3:a:oracle:mysql:5.1.26:::::::*|2019-12-17 15:07:17\n5.1.27|cpe:2.3:a:mysql:mysql:5.1.27:::::::*|cpe:2.3:a:oracle:mysql:5.1.27:::::::*|2019-12-17 15:07:24\n5.1.28|cpe:2.3:a:mysql:mysql:5.1.28:::::::*|cpe:2.3:a:oracle:mysql:5.1.28:::::::*|2019-12-17 15:07:31\n5.1.29|cpe:2.3:a:mysql:mysql:5.1.29:::::::*|cpe:2.3:a:oracle:mysql:5.1.29:::::::*|2019-12-17 15:07:38\n5.1.30|cpe:2.3:a:mysql:mysql:5.1.30:::::::*|cpe:2.3:a:oracle:mysql:5.1.30:::::::*|2019-12-17 15:07:46\n5.1.31:sp1|cpe:2.3:a:mysql:mysql:5.1.31:sp1::::::|cpe:2.3:a:oracle:mysql:5.1.31:sp1::::::|2019-12-17 15:23:35\n5.1.33|cpe:2.3:a:mysql:mysql:5.1.33:::::::*|cpe:2.3:a:oracle:mysql:5.1.33:::::::*|2019-12-17 15:07:55\n5.1.34:sp1|cpe:2.3:a:mysql:mysql:5.1.34:sp1::::::|cpe:2.3:a:oracle:mysql:5.1.34:sp1::::::|2019-12-17 15:23:42\n5.1.35|cpe:2.3:a:mysql:mysql:5.1.35:::::::*|cpe:2.3:a:oracle:mysql:5.1.35:::::::*|2019-12-17 15:08:03\n5.1.36|cpe:2.3:a:mysql:mysql:5.1.36:::::::*|cpe:2.3:a:oracle:mysql:5.1.36:::::::*|2019-12-17 15:08:10\n5.1.37:sp1|cpe:2.3:a:mysql:mysql:5.1.37:sp1::::::|cpe:2.3:a:oracle:mysql:5.1.37:sp1::::::|2019-12-17 15:23:49\n5.1.38|cpe:2.3:a:mysql:mysql:5.1.38:::::::*|cpe:2.3:a:oracle:mysql:5.1.38:::::::*|2019-12-17 15:08:17\n5.1.39|cpe:2.3:a:mysql:mysql:5.1.39:::::::*|cpe:2.3:a:oracle:mysql:5.1.39:::::::*|2019-12-17 10:22:31\n5.1.40|cpe:2.3:a:mysql:mysql:5.1.40:::::::*|cpe:2.3:a:oracle:mysql:5.1.40:::::::*|2019-12-17 10:22:46\n5.1.40:sp1|cpe:2.3:a:mysql:mysql:5.1.40:sp1::::::|cpe:2.3:a:oracle:mysql:5.1.40:sp1::::::|2019-12-17 10:22:39\n5.1.41|cpe:2.3:a:mysql:mysql:5.1.41:::::::*|cpe:2.3:a:oracle:mysql:5.1.41:::::::*|2019-12-17 10:22:52\n5.1.42|cpe:2.3:a:mysql:mysql:5.1.42:::::::*|cpe:2.3:a:oracle:mysql:5.1.42:::::::*|2019-12-17 10:22:59\n5.1.43|cpe:2.3:a:mysql:mysql:5.1.43:::::::*|cpe:2.3:a:oracle:mysql:5.1.43:::::::*|2019-12-17 10:23:06\n5.1.43:sp1|cpe:2.3:a:mysql:mysql:5.1.43:sp1::::::|cpe:2.3:a:oracle:mysql:5.1.43:sp1::::::|2019-12-17 10:23:11\n5.1.44|cpe:2.3:a:mysql:mysql:5.1.44:::::::*|cpe:2.3:a:oracle:mysql:5.1.44:::::::*|2019-12-17 10:23:17\n5.1.45|cpe:2.3:a:mysql:mysql:5.1.45:::::::*|cpe:2.3:a:oracle:mysql:5.1.45:::::::*|2019-12-17 10:23:24\n5.1.46|cpe:2.3:a:mysql:mysql:5.1.46:::::::*|cpe:2.3:a:oracle:mysql:5.1.46:::::::*|2019-12-17 10:23:38\n5.1.46:sp1|cpe:2.3:a:mysql:mysql:5.1.46:sp1::::::|cpe:2.3:a:oracle:mysql:5.1.46:sp1::::::|2019-12-17 10:23:31\n5.1.47|cpe:2.3:a:mysql:mysql:5.1.47:::::::*|cpe:2.3:a:oracle:mysql:5.1.47:::::::*|2019-12-17 10:23:45", "metadata": {"State-Gated Retrieval": ["Retain only CPE 2.3 entries in the MySQL 5.1 branch for which the vendor migration from mysql:mysql to oracle:mysql is clearly visible in the NVD Change History for all four vulnerabilities CVE-2010-3676, 3677, 3678, and 3680.", "Every result must provide version_update, legacy_cpe, current_cpe, and remap_time_nvd.", "Within the same version, entries without an update come before entries with an update, and output is sorted by numeric version in ascending order."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["Current and deprecated MySQL 5.1 CPE search results yield candidate legacy/current CPE strings", "The four target CVE pages and their Change History entries yield vendor-migration evidence, exact version_update strings, and remap timestamps", "Final rows are produced only after taking the common mysql:mysql 5.1 -> oracle:mysql 5.1 migrations across all four CVEs and removing alias-only or 5.0 entries"], "control_dependency": ["Legacy vendor mapping must be determined from Analysis Description plus Change History rather than from the current description/current CPE view", "Include-deprecated results cannot be accepted wholesale because alias-only remaps must be filtered out", "5.1-branch membership and sort order must be determined after the vendor-migration filter removes 5.0 and non-migration rows"], "freeze": {"historical_window": "The MySQL 5.1 branches visible in CVE-2010-3676, CVE-2010-3677, CVE-2010-3678, and CVE-2010-3680, restricted to explicit mysql:mysql -> oracle:mysql vendor-migration events"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Retain only CPE 2.3 entries in the MySQL 5.1 branch for which the vendor migration from mysql:mysql to oracle:mysql is clearly visible in the NVD Change History for all four vulnerabilities CVE-2010-3676, 3677, 3678, and 3680.", "Every result must provide version_update, legacy_cpe, current_cpe, and remap_time_nvd.", "Within the same version, entries without an update come before entries with an update, and output is sorted by numeric version in ascending order."], "exclusion_conditions": ["Exclude MySQL 5.0 branch entries.", "Exclude entries that still belong to the old vendor mysql:mysql and only underwent internal version alias normalization without a vendor migration.", "Exclude results that only read from the current Oracle MySQL 5.1 CPE state without using legacy mapping and Change History."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "Required literal first row `version_update|legacy_cpe|current_cpe|remap_time_nvd`", "schema": ["version_update", "legacy_cpe", "current_cpe", "remap_time_nvd"], "version_update": "Sort by numeric MySQL version ascending; within the same base version, the empty update comes before any non-empty update suffix", "legacy_cpe": "Preserve the exact legacy CPE 2.3 string", "current_cpe": "Preserve the exact current CPE 2.3 string", "remap_time_nvd": "Emit the NVD remap timestamp in `YYYY-MM-DD HH:MM:SS` form", "dedup_key": "(legacy_cpe, current_cpe)"}}}
{"task_id": "nvd_004-g", "domain": "NVD", "autonomy_type": "ordered table", "instruction": "Identify which milestone versions of Apache Tomcat 9.0.0 (prior to the final release) appear in the NVD current affected-product configurations for all four vulnerabilities CVE-2016-5018, CVE-2016-6794, CVE-2016-6796, and CVE-2016-6797, and whose NVD change history shows a CPE identifier migration from the legacy mN form to the current milestoneN form. This step is intended to clean up SBOM rules that still use the old Apache Tomcat 9.0.0.Mx notation. Here, mN refers to forms such as m1, m2, etc., while milestoneN refers to the corresponding notation currently used by NVD. Retain only those Apache Tomcat 9.0.0 milestone versions that satisfy both conditions. Exclude two categories: (1) milestone versions that do not simultaneously meet the four-CVE condition and appear only because of other Tomcat 9.0.0 prerelease vulnerabilities; (2) Tomcat 6, Tomcat 7, Tomcat 8, Tomcat 8.5, and any non-Tomcat products. Output a multi-row ordered table with columns milestone | legacy_cpe | current_cpe | required_cves | remap_time_nvd, sorted by the numeric milestone value in ascending order.", "start_url": "https://nvd.nist.gov/products/cpe/search", "output_format": "Output a multi-row ordered table with columns milestone | legacy_cpe | current_cpe | required_cves | remap_time_nvd, sorted by the numeric milestone value in ascending order.", "oracle_output_cardinality": 9, "oracle_answer": "milestone|legacy_cpe|current_cpe|required_cves|remap_time_nvd\nM1|cpe:2.3:a:apache:tomcat:9.0.0:m1:::::::*|cpe:2.3:a:apache:tomcat:9.0.0:milestone1:::::::*|CVE-2016-5018;CVE-2016-6794;CVE-2016-6796;CVE-2016-6797|2023-12-08 11:41:18\nM2|cpe:2.3:a:apache:tomcat:9.0.0:m2:::::::*|cpe:2.3:a:apache:tomcat:9.0.0:milestone2:::::::*|CVE-2016-5018;CVE-2016-6794;CVE-2016-6796;CVE-2016-6797|2023-12-08 11:41:18\nM3|cpe:2.3:a:apache:tomcat:9.0.0:m3:::::::*|cpe:2.3:a:apache:tomcat:9.0.0:milestone3:::::::*|CVE-2016-5018;CVE-2016-6794;CVE-2016-6796;CVE-2016-6797|2023-12-08 11:41:18\nM4|cpe:2.3:a:apache:tomcat:9.0.0:m4:::::::*|cpe:2.3:a:apache:tomcat:9.0.0:milestone4:::::::*|CVE-2016-5018;CVE-2016-6794;CVE-2016-6796;CVE-2016-6797|2023-12-08 11:41:18\nM5|cpe:2.3:a:apache:tomcat:9.0.0:m5:::::::*|cpe:2.3:a:apache:tomcat:9.0.0:milestone5:::::::*|CVE-2016-5018;CVE-2016-6794;CVE-2016-6796;CVE-2016-6797|2023-12-08 11:41:18\nM6|cpe:2.3:a:apache:tomcat:9.0.0:m6:::::::*|cpe:2.3:a:apache:tomcat:9.0.0:milestone6:::::::*|CVE-2016-5018;CVE-2016-6794;CVE-2016-6796;CVE-2016-6797|2023-12-08 11:41:18\nM7|cpe:2.3:a:apache:tomcat:9.0.0:m7:::::::*|cpe:2.3:a:apache:tomcat:9.0.0:milestone7:::::::*|CVE-2016-5018;CVE-2016-6794;CVE-2016-6796;CVE-2016-6797|2023-12-08 11:41:18\nM8|cpe:2.3:a:apache:tomcat:9.0.0:m8:::::::*|cpe:2.3:a:apache:tomcat:9.0.0:milestone8:::::::*|CVE-2016-5018;CVE-2016-6794;CVE-2016-6796;CVE-2016-6797|2023-12-08 11:41:18\nM9|cpe:2.3:a:apache:tomcat:9.0.0:m9:::::::*|cpe:2.3:a:apache:tomcat:9.0.0:milestone9:::::::*|CVE-2016-5018;CVE-2016-6794;CVE-2016-6796;CVE-2016-6797|2023-12-08 11:41:18", "metadata": {"State-Gated Retrieval": ["Retain only Apache Tomcat 9.0.0 milestone versions that appear in the current affected-product scope of all four target CVEs: CVE-2016-5018, CVE-2016-6794, CVE-2016-6796, and CVE-2016-6797.", "These versions must also show a CPE Deprecation Remap from mN to milestoneN in the NVD Change History.", "Finally, output sorted by numeric milestone value in ascending order, and the required_cves field must contain exactly these four specified CVEs."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the current milestoneN CPE search and detail pages yield the Tomcat milestone family and the legacy mN mapping anchors", "the four target CVE pages and Change History entries yield the exact mN -> milestoneN deprecation remaps and the required-CVE intersection", "final rows are produced only after taking the remaps common to all four target CVEs and sorting by numeric milestone"], "control_dependency": ["the candidate milestone set must be determined from the four-target-CVE intersection rather than from all prerelease Tomcat CVEs", "later remapped milestones such as M10+ must be excluded by applying the specified four-CVE boundary", "only explicit NVD CPE Deprecation Remap events from mN to milestoneN count, so generic Tomcat prerelease visibility cannot be accepted as evidence"], "freeze": {"historical_window": "the Tomcat 9.0.0 prerelease milestones jointly covered by CVE-2016-5018, CVE-2016-6794, CVE-2016-6796, and CVE-2016-6797, with remap evidence fixed to the 2023-12-08 11:41:18 NVD deprecation events"}, "answer_type": "multi-row ordered table"}, "all_involved_urls": {"fixed_entry_and_docs": ["https://nvd.nist.gov/products/cpe/search", "https://nvd.nist.gov/vuln/vulnerability-detail-pages"], "cpe_search_result_pages": ["https://nvd.nist.gov/products/cpe/search/results?keyword=apache%20tomcat%209.0.0%20milestone&namingFormat=2.3&orderBy=CPEURI&status=FINAL", "https://nvd.nist.gov/products/cpe/search/results?keyword=apache%20tomcat%209.0.0%20milestone&namingFormat=2.3&orderBy=CPEURI&status=FINAL%2CDEPRECATED", "https://nvd.nist.gov/products/cpe/search/results?keyword=apache%20tomcat%209.0.0%20m&namingFormat=2.3&orderBy=CPEURI&status=FINAL%2CDEPRECATED", "https://nvd.nist.gov/products/cpe/search/results?keyword=cpe%3A2.3%3Aa%3Aapache%3Atomcat%3A9.0.0%3Amilestone1%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A&namingFormat=2.3&orderBy=CPEURI&status=FINAL%2CDEPRECATED"], "representative_cpe_detail_pages": ["https://nvd.nist.gov/products/cpe/detail/615922?keyword=cpe%3A2.3%3Aa%3Aapache%3Atomcat&namingFormat=2.3&orderBy=CPEURI&status=FINAL%2CDEPRECATED", "https://nvd.nist.gov/products/cpe/detail/615947?keyword=cpe%3A2.3%3Aa%3Aapache&namingFormat=2.3&orderBy=CPEURI&status=FINAL%2CDEPRECATED"], "row_level_cpe_api_urls_m1_to_m9": ["https://services.nvd.nist.gov/rest/json/cpes/2.0?cpeMatchString=cpe%3A2.3%3Aa%3Aapache%3Atomcat%3A9.0.0%3Amilestone1%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A", "https://services.nvd.nist.gov/rest/json/cpes/2.0?cpeMatchString=cpe%3A2.3%3Aa%3Aapache%3Atomcat%3A9.0.0%3Amilestone2%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A", "https://services.nvd.nist.gov/rest/json/cpes/2.0?cpeMatchString=cpe%3A2.3%3Aa%3Aapache%3Atomcat%3A9.0.0%3Amilestone3%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A", "https://services.nvd.nist.gov/rest/json/cpes/2.0?cpeMatchString=cpe%3A2.3%3Aa%3Aapache%3Atomcat%3A9.0.0%3Amilestone4%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A", "https://services.nvd.nist.gov/rest/json/cpes/2.0?cpeMatchString=cpe%3A2.3%3Aa%3Aapache%3Atomcat%3A9.0.0%3Amilestone5%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A", "https://services.nvd.nist.gov/rest/json/cpes/2.0?cpeMatchString=cpe%3A2.3%3Aa%3Aapache%3Atomcat%3A9.0.0%3Amilestone6%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A", "https://services.nvd.nist.gov/rest/json/cpes/2.0?cpeMatchString=cpe%3A2.3%3Aa%3Aapache%3Atomcat%3A9.0.0%3Amilestone7%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A", "https://services.nvd.nist.gov/rest/json/cpes/2.0?cpeMatchString=cpe%3A2.3%3Aa%3Aapache%3Atomcat%3A9.0.0%3Amilestone8%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A", "https://services.nvd.nist.gov/rest/json/cpes/2.0?cpeMatchString=cpe%3A2.3%3Aa%3Aapache%3Atomcat%3A9.0.0%3Amilestone9%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A%3A%2A"], "target_cve_detail_pages": ["https://nvd.nist.gov/vuln/detail/CVE-2016-5018", "https://nvd.nist.gov/vuln/detail/CVE-2016-6794", "https://nvd.nist.gov/vuln/detail/CVE-2016-6796", "https://nvd.nist.gov/vuln/detail/CVE-2016-6797"], "negative_control_cve_detail_pages": ["https://nvd.nist.gov/vuln/detail/CVE-2016-6816", "https://nvd.nist.gov/vuln/detail/CVE-2016-6817", "https://nvd.nist.gov/vuln/detail/CVE-2019-0221"], "cve_change_history_api": ["https://services.nvd.nist.gov/rest/json/cvehistory/2.0?cveId=CVE-2016-5018&eventName=CPE%20Deprecation%20Remap", "https://services.nvd.nist.gov/rest/json/cvehistory/2.0?cveId=CVE-2016-6794&eventName=CPE%20Deprecation%20Remap", "https://services.nvd.nist.gov/rest/json/cvehistory/2.0?cveId=CVE-2016-6796&eventName=CPE%20Deprecation%20Remap", "https://services.nvd.nist.gov/rest/json/cvehistory/2.0?cveId=CVE-2016-6797&eventName=CPE%20Deprecation%20Remap", "https://services.nvd.nist.gov/rest/json/cvehistory/2.0?cveId=CVE-2016-6816&eventName=CPE%20Deprecation%20Remap", "https://services.nvd.nist.gov/rest/json/cvehistory/2.0?cveId=CVE-2016-6817&eventName=CPE%20Deprecation%20Remap", "https://services.nvd.nist.gov/rest/json/cvehistory/2.0?cveId=CVE-2019-0221&eventName=CPE%20Deprecation%20Remap"], "representative_change_record_pages": ["https://nvd.nist.gov/vuln/detail/CVE-2016-5018/change-record?changeRecordedOn=04%2F18%2F2022T13%3A57%3A03.500-0400", "https://nvd.nist.gov/vuln/detail/CVE-2016-5018/change-record?changeRecordedOn=08%2F24%2F2017T09%3A11%3A40.690-0400"]}, "rubric": {"inclusion_conditions": ["Retain only Apache Tomcat 9.0.0 milestone versions that appear in the current affected-product scope of all four target CVEs: CVE-2016-5018, CVE-2016-6794, CVE-2016-6796, and CVE-2016-6797.", "These versions must also show a CPE Deprecation Remap from mN to milestoneN in the NVD Change History.", "Finally, output sorted by numeric milestone value in ascending order, and the required_cves field must contain exactly these four specified CVEs."], "exclusion_conditions": ["Exclude milestones that are included only because of other Tomcat 9.0.0 prerelease vulnerabilities, such as M10/M11 and later milestones.", "Exclude Tomcat 6, 7, 8, 8.5, and any non-Tomcat products.", "Exclude records that lack explicit NVD remap evidence and are merely visible on the current milestone page."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "required literal first row `milestone|legacy_cpe|current_cpe|required_cves|remap_time_nvd`", "schema": ["milestone", "legacy_cpe", "current_cpe", "required_cves", "remap_time_nvd"], "milestone": "emit the canonical milestone label `M1` through `M9` and sort by the numeric suffix ascending", "legacy_cpe": "preserve the exact legacy CPE 2.3 string containing the short milestone token such as `m1`", "current_cpe": "preserve the exact current CPE 2.3 string containing the full milestone token such as `milestone1`", "required_cves": "emit the fixed literal `CVE-2016-5018;CVE-2016-6794;CVE-2016-6796;CVE-2016-6797`", "remap_time_nvd": "emit the NVD remap timestamp in `YYYY-MM-DD HH:MM:SS` form"}}}
{"task_id": "reptile_001-g", "domain": "REPTILE_DATABASE", "autonomy_type": "ordered table", "oracle_output_cardinality": 8, "instruction": "I am compiling a name concordance for early museum specimens of Indian snakes recorded in George A. Boulenger's 1890 work \"The Fauna of British India, Including Ceylon and Burma\". In the Reptile Database, identify entries that simultaneously satisfy the following conditions: the currently accepted name has been reassigned to a different genus compared with the name used in the book; the current species page still retains BMNH type specimen information; and the type locality still corresponds to India. For each qualifying entry, also retrieve the most recent historical binomial listed on the current species page that has not yet adopted the current genus name, and determine whether an exact search using that full old name uniquely identifies the entry. Output the results sorted alphabetically by accepted name in the format <accepted name>|<old label>|<UNIQUE/AMBIG>|<BMNH status>; join multiple entries with semicolons; if no qualifying entries exist, output NONE.", "start_url": "https://reptile-database.reptarium.cz/advanced_search", "output_format": "Output the results sorted alphabetically by accepted name in the format <accepted name>|<old label>|<UNIQUE/AMBIG>|<BMNH status>; join multiple entries with semicolons; if no qualifying entries exist, output NONE.", "oracle_answer": "Argyrophis oatesii|Typhlops oatesii|UNIQUE|SYNTYPES;Dendrelaphis grandoculis|Ahaetulla grandoculis|UNIQUE|LECTOTYPE;Fowlea sanctijohannis|Xenochrophis sanctijohannis|UNIQUE|HOLOTYPE;Gerrhopilus beddomii|Typhlops beddomii|UNIQUE|SYNTYPES;Hebius khasiensis|Amphiesma khasiense|UNIQUE|LECTOTYPE;Hebius parallelus|Amphiesma parallelum|AMBIG|LECTOTYPE;Indotyphlops jerdoni|Typhlops jerdoni|UNIQUE|SYNTYPES;Myriopholis blanfordii|Leptotyphlops blanfordi|AMBIG|SYNTYPES", "metadata": {"State-Gated Retrieval": ["Candidates must come from the historical Indian snake collection entries in Boulenger 1890 \"Fauna of British India\".", "The current Reptile Database species page must simultaneously satisfy: the species has been reassigned to a different genus; the page still retains BMNH type specimen information; and the type locality still corresponds to India.", "On each qualifying page, select the most recent historical binomial that has not yet adopted the current genus name, perform an exact back-search using that old name, and then label it UNIQUE or AMBIG."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["The Boulenger 1890 historical sample and current accepted-species pages yield candidates that changed genus while retaining BMNH type data and India-compatible type locality.", "Each accepted page yields the nearest pre-current-genus old label, the BMNH type status, and the historical/type-locality evidence anchoring the entity.", "Final rows are produced only after the old-label exact back-search is evaluated for uniqueness without losing the original entity anchor."], "control_dependency": ["When an exact old-label back-search now hits multiple species, the original anchored entity must be kept and the uniqueness label must be set to AMBIG.", "Once a candidate has passed the historical-cluster, genus-shift, BMNH, and India-locality gates, later old-label ambiguity can only rewrite the uniqueness field, not the entity itself.", "Candidate anchoring must be preserved across the two-stage workflow so that old-label search noise does not redirect the accepted-species row."], "freeze": {"historical_window": "The objectively existing reptile taxa in the George A. Boulenger 1890 British India sample, using Reptile Database accepted pages, synonym labels, BMNH type data, and type-locality text as nearly objective and highly stable reference information."}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Candidates must come from the historical Indian snake collection entries in Boulenger 1890 \"Fauna of British India\".", "The current Reptile Database species page must simultaneously satisfy: the species has been reassigned to a different genus; the page still retains BMNH type specimen information; and the type locality still corresponds to India.", "On each qualifying page, select the most recent historical binomial that has not yet adopted the current genus name, perform an exact back-search using that old name, and then label it UNIQUE or AMBIG."], "exclusion_conditions": ["Exclude entries where the current accepted name has not undergone a genus change, or where the page no longer retains BMNH type specimen information.", "Exclude entries whose type locality does not correspond to India.", "Do not discard candidates for which the exact old-name back-search is non-unique; such candidates should retain the original entity and be labeled AMBIG."], "normalization": {"field_separator": "|", "record_separator": ";", "schema": ["accepted name", "old label", "UNIQUE/AMBIG", "BMNH status"], "accepted_name": "Emit the current accepted binomial exactly as shown in the reconciled source.", "old_label": "Preserve the historical label exactly as used on the legacy branch.", "UNIQUE/AMBIG": "Emit `UNIQUE` or `AMBIG` in uppercase.", "BMNH_status": "Normalize the type-status field to uppercase labels such as HOLOTYPE, LECTOTYPE, or SYNTYPES.", "sorting_or_selection": "Accepted name alphabetical ascending."}}}
{"task_id": "reptile_002-g", "domain": "REPTILE_DATABASE", "autonomy_type": "ordered table", "oracle_output_cardinality": 3, "instruction": "I am compiling name correspondences between old labels and current accepted names for early museum specimens of Southeast Asian Lygosoma. In the current Lygosoma checklist, identify all accepted species that satisfy both of the following conditions: the species page lists a Riopa combination used by Smith (1935/1937) in its Synonym section, and that old name corresponds to more than one accepted species under the current classification. For each qualifying species, output the first-listed type specimen number, type status, and the region corresponding to the final type locality on that page. Output in alphabetical order by accepted name, using the format <accepted>|<Smith-era Riopa label>|<first type code>|<type status>|<region>; join multiple entries with semicolons. If no qualifying items exist, output NONE.", "start_url": "https://reptile-database.reptarium.cz/advanced_search", "output_format": "Output in alphabetical order by accepted name, using the format <accepted>|<Smith-era Riopa label>|<first type code>|<type status>|<region>; join multiple entries with semicolons. If no qualifying items exist, output NONE.", "oracle_answer": "Lygosoma bampfyldei|Riopa bampfyldei|BMNH 1946.8.10.84|SYNTYPES|East Malaysia;Lygosoma peninsulare|Riopa bampfyldei|LSUHC 13857|HOLOTYPE|Peninsular Malaysia;Lygosoma schneideri|Riopa bampfyldei|NMBA 4743|HOLOTYPE|Indonesia", "metadata": {"State-Gated Retrieval": ["Retain only accepted species that can be traced from the Smith 1937 old-name chain to a current Lygosoma record and for which the type accession, type status, and locality can be read on the current page.", "For each accepted record, output the accepted name, the corresponding old label, the type accession, the type status, and the locality.", "If the same exact old name appears on multiple accepted species, retain them as separate chains."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the Smith 1937 old-name chain yields the three current Lygosoma accepted pages connected to Riopa bampfyldei / Lygosoma bampfyldei", "each accepted page yields its own type accession, type status, and type locality needed to separate the homonymous old-name paths", "final rows are produced only after the one-to-many old-name split is rebuilt and the three paths are re-anchored to their own type evidence"], "control_dependency": ["same-specific-epithet matching cannot establish a one-to-one mapping when the same Smith 1937 old name appears on multiple accepted pages", "once literal old-name reuse is exposed, the workflow must keep the original parallel three-chain state rather than collapsing to a one-to-one mapping", "type accession and locality evidence must be read separately for each chain rather than patched onto the first stopping page"], "freeze": {"historical_window": "the objectively existing taxa linked to the Smith 1937 old-name cluster around Riopa bampfyldei, using Reptile Database accepted pages together with type accessions, statuses, and localities as nearly objective and highly stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Retain only accepted species that can be traced from the Smith 1937 old-name chain to a current Lygosoma record and for which the type accession, type status, and locality can be read on the current page.", "For each accepted record, output the accepted name, the corresponding old label, the type accession, the type status, and the locality.", "If the same exact old name appears on multiple accepted species, retain them as separate chains."], "exclusion_conditions": ["Exclude results that assume the same species simply because the old name and current specific epithet match.", "Exclude results that retain only the first chain when the same exact old name appears on multiple accepted pages.", "Exclude results that do not re-verify the type accession and locality on each accepted page."], "normalization": {"field_separator": "|", "record_separator": ";", "schema": ["accepted", "Smith-era Riopa label", "first type code", "type status", "region"], "accepted": "emit the current accepted binomial exactly as shown in the reconciled source", "Smith-era_Riopa_label": "preserve the Smith-era Riopa label exactly as used on the legacy branch", "first_type_code": "emit the first accession code that follows the selected type-status statement", "type_status": "normalize the type-status field to uppercase labels such as HOLOTYPE or SYNTYPES", "region": "emit the modern geographic region label from the source reconciliation, such as East Malaysia or Peninsular Malaysia", "sorting_or_selection": "accepted name alphabetical ascending"}}}
{"task_id": "reptile_003-g", "domain": "REPTILE_DATABASE", "autonomy_type": "ordered table", "oracle_output_cardinality": 10, "instruction": "I am compiling a name cross-reference for a batch of old collection cards that were still cataloged under the names Mabuya or Mabuia around 1980 and now need to be merged into the current Eutropis checklist. In the current Eutropis list, identify all entries that meet the following conditions: the first 1980-era label on the species page has the genus written as Mabuya or Mabuia; and when that 1980-era label is searched by name in the current list, it does not uniquely correspond to a single Eutropis entry but instead matches multiple current Eutropis entries. For each qualifying entry, output the current accepted name, that 1980-era label, the qualifier for that label on the page, and the first type status and type code listed on the current page. Output sorted alphabetically by 1980-era label, and within the same label by current accepted name, in the format <accepted>|<1980 label>|<qualifier>|<type status>|<first type code>; join multiple entries with semicolons; if no qualifying entries exist, output NONE.", "start_url": "https://reptile-database.reptarium.cz/advanced_search", "output_format": "Output sorted alphabetically by 1980-era label, and within the same label by current accepted name, in the format <accepted>|<1980 label>|<qualifier>|<type status>|<first type code>; join multiple entries with semicolons; if no qualifying entries exist, output NONE.", "oracle_answer": "Eutropis indeprensa|Mabuya indeprensa|NONE|HOLOTYPE|CAS 85664;Eutropis sahulinghangganan|Mabuya indeprensa|PART-MISID|HOLOTYPE|PNM 9867;Eutropis multifasciata|Mabuya multifasciata|NONE|NEOTYPE|MZB 11912;Eutropis multicarinata|Mabuya multicarinata|NONE|HOLOTYPE|BMNH 1946.8.15.13-15;Eutropis borealis|Mabuya multicarinata borealis|NONE|HOLOTYPE|CAS 15447;Eutropis islamaliit|Mabuya multicarinata borealis|PART-MISID|HOLOTYPE|PNM 9847;Eutropis caraga|Mabuya multicarinata multicarinata|PART-MISID|HOLOTYPE|PNM 9845;Eutropis palauensis|Mabuya multicarinata multicarinata|PART-BYIMPL|HOLOTYPE|CAS 248251;Eutropis dattaroyi|Mabuya rudis|PART|HOLOTYPE|ZSI 25118;Eutropis rudis|Mabuya rudis|NONE|LECTOTYPE|BMNH 1946.8.15.26", "metadata": {"State-Gated Retrieval": ["Retain only those current Eutropis records whose synonym block contains, scanning top-down, a first old label starting with Mabuya or Mabuia and having the year 1980.", "Then use that 1980-era old label for a default name back-search, and output the accepted name, old label, back-search qualifier, type status, and type specimen.", "If the subsequent back-search reveals additional matching entries, revisit the earlier results and include them in the final output."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["current Eutropis detail pages yield the first 1980-tagged Mabuya/Mabuia synonym label for each accepted species", "exact back-search on that 1980 old label yields the present accepted page, type status, and type specimen needed for the output row", "final rows are produced only after the 1980-specific old-label anchor is rebuilt and the accepted-species set is expanded when back-search evidence reveals omitted entries"], "control_dependency": ["the workflow must identify the first Mabuya/Mabuia label carrying the 1980 year rather than stopping at the first old label on the page", "when old-label back-search reveals missing entries such as Eutropis dattaroyi, the accepted-species set must be expanded to include them rather than patched in place", "accepted-page synonym parsing and back-search interpretation must stay coupled so that 1980 labels are not replaced by older but irrelevant names"], "freeze": {"historical_window": "the objectively existing Eutropis taxa represented in Reptile Database pages, using the synonym labels shown on each species page, especially the first Mabuya/Mabuia-form label tagged with the 1980 citation year, as nearly objective and highly stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Retain only those current Eutropis records whose synonym block contains, scanning top-down, a first old label starting with Mabuya or Mabuia and having the year 1980.", "Then use that 1980-era old label for a default name back-search, and output the accepted name, old label, back-search qualifier, type status, and type specimen.", "If the subsequent back-search reveals additional matching entries, revisit the earlier results and include them in the final output."], "exclusion_conditions": ["Exclude results that mistakenly treat the first old label on the page as the first 1980-era old label.", "Exclude results whose exact back-search on the 1980 old label points to only a single current Eutropis entry.", "Exclude results that, when subsequent evidence reveals omitted entries, fail to revisit the earlier results and rebuild the accepted-species set."], "normalization": {"field_separator": "|", "record_separator": ";", "schema": ["accepted", "1980 label", "qualifier", "type status", "first type code"], "accepted": "emit the current accepted binomial exactly as shown in the reconciled source", "1980_label": "preserve the 1980-era label exactly as used on the legacy branch", "qualifier": "normalize to one of `NONE`, `PART`, `PART-MISID`, or `PART-BYIMPL`", "type_status": "normalize the type-status field to uppercase labels such as HOLOTYPE or SYNTYPES", "first_type_code": "emit the first accession code that follows the selected type-status statement", "sorting_or_selection": {"primary": "1980 label alphabetical ascending", "secondary": "accepted name alphabetical ascending"}}}}
{"task_id": "reptile_004-g", "domain": "REPTILE_DATABASE", "autonomy_type": "ordered table", "oracle_output_cardinality": 6, "instruction": "I am organizing a set of old Hemidactylus brookii museum cards that use the historical labels listed in the comments of the Hemidactylus brookii species page. Identify all old labels that are no longer considered synonyms of Hemidactylus brookii and for which the database provides a clear final synonym assignment. For each qualifying old label, output the currently accepted species it should be assigned to; if the synonym chain passes through an intermediate historical name, list that name, otherwise output DIRECT; also output the type status, type code, and the country of the type locality for that old label. Output the results in alphabetical order of the old labels, using the format <legacy label>|<accepted>|<bridge>|<type status>|<type code>|<country>, with multiple entries separated by semicolons. If no qualifying entries exist, output NONE.", "start_url": "https://reptile-database.reptarium.cz/advanced_search", "output_format": "Output in alphabetical order of the old labels, using the format <legacy label>|<accepted>|<bridge>|<type status>|<type code>|<country>, with multiple entries separated by semicolons. If no qualifying entries exist, output NONE.", "oracle_answer": "Hemidactylus brooki leightoni|Hemidactylus angulatus|DIRECT|HOLOTYPE|BMNH 1946.8.25.65|Colombia;Hemidactylus brookii haitianus|Hemidactylus angulatus|DIRECT|LECTOTYPE|ZMH R22378|Haiti;Hemidactylus brookii parvimaculatus|Hemidactylus parvimaculatus|DIRECT|HOLOTYPE|NMSL No. RG. 15|Sri Lanka;Hemidactylus brookii subtriedroides|Hemidactylus murrayi|Hemidactylus tenkatei|LECTOTYPE|BMNH [subtriedroides] [tenkatei]|Myanmar;Hemidactylus gleadowi|Hemidactylus gleadowi|DIRECT|NEOTYPE|BMNH 1884.7.25.8|Pakistan;Hemidactylus tenkatei|Hemidactylus murrayi|DIRECT|LECTOTYPE|RMNH 4353|Indonesia", "metadata": {"State-Gated Retrieval": ["Each result must start from the specified old-label chain and ultimately determine the accepted name, bridge status, type status, type code, and locality from the current accepted species page.", "For the subtriedroides branch, continue the chain through tenkatei to murrayi, output the final accepted name as Hemidactylus murrayi, and retain bridge=Hemidactylus tenkatei.", "For branches such as haitianus and leightoni, read the locality and type lines specific to each branch, rather than borrowing the header information from the wrong accepted page."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the old-label chains from Hemidactylus angulatus and related names yield accepted pages, bridge names, and the type/locality lines for each path", "comment, Types, and type-comment blocks yield the later evidence that tenkatei is further sunk into H. murrayi", "final rows are produced only after each old-label path is re-anchored to its own locality/type line and the subtriedroides -> tenkatei -> murrayi bridge chain is completed"], "control_dependency": ["type data must be taken from the locality/type line belonging to the actual old-label branch rather than from the header block of the wrong accepted page", "stopping at the intermediate tenkatei bridge is insufficient; the workflow must continue to the later murrayi evidence and rewrite both tenkatei and subtriedroides final accepted names", "bridge labels remain useful, but final accepted names and type lines must be determined from the deeper synonym chain"], "freeze": {"historical_window": "the objectively existing taxa in the Hemidactylus angulatus / brookii / murrayi complex, using Reptile Database locality, type, comment, and type-comment blocks as nearly objective and highly stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Each result must start from the specified old-label chain and ultimately determine the accepted name, bridge status, type status, type code, and locality from the current accepted species page.", "For the subtriedroides branch, continue the chain through tenkatei to murrayi, output the final accepted name as Hemidactylus murrayi, and retain bridge=Hemidactylus tenkatei.", "For branches such as haitianus and leightoni, read the locality and type lines specific to each branch, rather than borrowing the header information from the wrong accepted page."], "exclusion_conditions": ["Exclude results that follow the wrong page hierarchy and apply the Holotype/locality from the H. angulatus page header to other branches.", "Exclude results that stop early on the subtriedroides -> tenkatei chain without continuing to read the subsequent murrayi evidence.", "Exclude results that fail to align the locality/type lines with the specific old-label branches one by one."], "normalization": {"field_separator": "|", "record_separator": ";", "schema": ["legacy label", "accepted", "bridge", "type status", "type code", "country"], "legacy_label": "preserve the legacy label exactly as used on the source branch", "accepted": "emit the current accepted binomial exactly as shown in the reconciled source", "bridge": "emit `DIRECT` when no intermediate legacy bridge name is needed; otherwise emit the intermediate legacy name that links the branches", "type_status": "normalize the type-status field to uppercase labels such as HOLOTYPE, LECTOTYPE, or NEOTYPE", "type_code": "take the type accession from the legacy-label branch rather than from a generic type line elsewhere on the page", "country": "emit the modern English country name from the reconciled source branch", "sorting_or_selection": "legacy label alphabetical ascending"}}}
{"task_id": "reptile_005-g", "domain": "REPTILE_DATABASE", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I want to organize this synonymization record for an old Sri Lankan museum name in the current Naja list into a two-row comparison table. Output two rows with columns rank|name_role|name|status_or_evidence|specimen_role|specimen_code|supporting_note. The first row should be ACCEPTED_NAME and the second row CONTESTED_NAME.", "start_url": "https://reptile-database.reptarium.cz/advanced_search?genus=Naja&submit=Search", "output_format": "Output a two-row comparison table with columns rank|name_role|name|status_or_evidence|specimen_role|specimen_code|supporting_note.", "oracle_answer": "1|ACCEPTED_NAME|Naja naja|SUBJECTIVE_JUNIOR_SYNONYM_OF_NAJA_NAJA|VALID_LECTOTYPE|ZMB 2796|restricted type locality: Sri Lanka\n2|CONTESTED_NAME|Naja polyocellata|SHI et al. 2022+SILVA et al. 2023|HOLOTYPE|BMNH 1946.1.18.50|invalid lectotype claim: NRM (= NHR) Lin-90", "metadata": {"State-Gated Retrieval": ["Within the current Naja list, identify the Sri Lanka contested-name branch: the name was recently reused as a distinct valid name, but the database now explicitly merges it back into the accepted name.", "On the current Naja naja page, read the recent-use evidence, final synonym status, lectotype correction, restricted type locality, and the holotype information for the contested name itself.", "The final output must organize the ACCEPTED_NAME side and the CONTESTED_NAME side into a two-row comparison table."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the current Naja species list and the Naja naja page yield the Sri Lanka contested-name branch, recent-use citations, and the accepted-name context", "comment, Types, and type-comment blocks yield the final synonym treatment plus the valid and invalid lectotype assignments and the contested-name holotype", "final rows are produced only after the accepted-name side and the contested-name side are reconciled on the current Naja naja page"], "control_dependency": ["recent-use evidence for Naja polyocellata must be interpreted together with the current final synonym treatment shown on that page", "lectotype status must be fixed from the combined Types and type-comment evidence rather than from a single provisional assignment", "the accepted-name and contested-name rows are emitted together only after final status, type locality, and specimen roles are fixed"], "freeze": {"historical_window": "the objectively existing taxa in the Sri Lanka Naja branch, using current Reptile Database comment, Types, and type-comment blocks as nearly objective and highly stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Within the current Naja list, identify the Sri Lanka contested-name record: the name was recently reused in the literature as a distinct valid name, but the database now explicitly states that it should still be merged into the accepted name.", "On the current Naja naja page, the result must jointly recover the recent-use evidence, final synonym status, lectotype correction, restricted type locality, and the holotype information for the contested name itself.", "The final output must organize the ACCEPTED_NAME side and the CONTESTED_NAME side into a two-row comparison table."], "exclusion_conditions": ["Exclude results that treat the recently revived contested name as the current accepted name.", "Exclude results that stop at the Wallach 2014 lectotype designation and do not continue to the later comment and type-comment correction.", "Exclude results that fail to output both the restricted type locality and the contested-name holotype."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["rank", "name_role", "name", "status_or_evidence", "specimen_role", "specimen_code", "supporting_note"], "rank": "emit 1 for ACCEPTED_NAME and 2 for CONTESTED_NAME", "name_role": ["ACCEPTED_NAME", "CONTESTED_NAME"], "name_fields": "use the current binomial spellings exactly", "specimen_code_normalization": "preserve museum abbreviations, spaces, parentheses, and specimen punctuation exactly as the page shows", "sorting_or_selection": "emit rows in the fixed order ACCEPTED_NAME then CONTESTED_NAME"}}}
{"task_id": "wateroffice_001-g", "domain": "WATER_OFFICE", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I am compiling historical daily flow comparison data for the Saint John River spring freshet and want to identify, along the Saint John River mainstem, the last station in the current monitoring network that still allows daily flow comparisons for March-April 1973 and March-April 2018, as well as the first downstream mainstem station that no longer satisfies that condition. Mark a station as PASS if the target windows have complete daily flow records, LEVEL_ONLY if only water level records are available, and FLOW_GAP if daily flow records exist but the target months are missing or incomplete. Output a two-row breakpoint table with columns rank|station_number|breakpoint_role|target_window_status|final_verdict. The first row should be LAST_USABLE and the second row should be FIRST_DOWNSTREAM_FAIL.", "start_url": "https://wateroffice.ec.gc.ca/search/real_time_e.html", "output_format": "Output a two-row breakpoint table with columns rank|station_number|breakpoint_role|target_window_status|final_verdict.", "oracle_answer": "1|01AF002|LAST_USABLE|PASS|KEEP\n2|01AK003|FIRST_DOWNSTREAM_FAIL|LEVEL_ONLY|REJECT", "metadata": {"State-Gated Retrieval": ["Determine the last station still usable for the target historical-window comparison only within the Saint John River mainstem stations in the current monitoring network.", "The historical availability page must distinguish Flow Data Availability from Daily Water Level Data Availability, and completeness for March-April 1973 and March-April 2018 must be checked.", "The final output must provide both breakpoint rows, LAST_USABLE and FIRST_DOWNSTREAM_FAIL, under the same validation rule."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the current Saint John River mainstem station sequence provides the downstream order in which the breakpoints must be checked", "per-station historical availability pages yield Flow Data Availability, Daily Water Level Data Availability, and target-month completeness for 1973 and 2018", "the final breakpoint rows are produced only after the availability-page checks identify the last usable station and the first downstream failure under the same rule"], "control_dependency": ["the breakpoint pair must be determined from historical availability that distinguishes flow from level-only evidence", "target-year/month completeness must be checked on the availability page rather than inferred from the station list", "the second row is emitted only after the first downstream failing mainstem station is fixed under the same validation rule"], "freeze": {"historical_window": "mainstem candidate stations evaluated within the fixed historical March/April daily-availability windows for 1973 and 2018, using station ordering and completeness signals as nearly stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Determine the last station still usable for the target historical-window comparison only within the Saint John River mainstem stations in the current monitoring network.", "The historical availability page must distinguish Flow Data Availability from Daily Water Level Data Availability, and completeness for March-April 1973 and March-April 2018 must be checked.", "The final output must provide both breakpoint rows, LAST_USABLE and FIRST_DOWNSTREAM_FAIL, under the same validation rule."], "exclusion_conditions": ["Exclude results that determine the breakpoint directly from the current Saint John River mainstem station sequence using only a reference index.", "Exclude results that mistakenly treat level-only evidence as satisfying the flow condition.", "Exclude results that do not check each station's availability page to verify target-month completeness."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["rank", "station_number", "breakpoint_role", "target_window_status", "final_verdict"], "rank": "emit `1` for LAST_USABLE and `2` for FIRST_DOWNSTREAM_FAIL", "station_number": "preserve the station number exactly", "breakpoint_role": "emit `LAST_USABLE` and `FIRST_DOWNSTREAM_FAIL` exactly", "target_window_status": "emit `PASS`, `LEVEL_ONLY`, or `FLOW_GAP` exactly", "final_verdict": "emit `KEEP` for the usable breakpoint row and `REJECT` for the downstream failing breakpoint row", "stop_condition": ["the last usable mainstem station is fixed under the availability-page validation rule", "the first downstream failing mainstem station is fixed under the same validation rule before the two-row output is emitted"]}}}
{"task_id": "wateroffice_002-g", "domain": "WATER_OFFICE", "autonomy_type": "ordered table", "oracle_output_cardinality": 4, "instruction": "I am compiling historical daily flow records for the Bulkley River reach from Houston to Hazelton and want to determine which candidate mainstem stations on this reach truly have complete daily flow records for April-July 1948 and April-July 1950. Mark a station as PASS if the target windows have complete daily flow records, LEVEL_ONLY if only water level records are available, and FLOW_GAP if daily flow records exist but any target month from April to July is incomplete. Output a table in downstream-to-upstream order with columns rank|station_number|target_window_status|final_verdict.", "start_url": "https://wateroffice.ec.gc.ca/search/historical_e.html", "output_format": "Output a table in downstream-to-upstream order with columns rank|station_number|target_window_status|final_verdict.", "oracle_answer": "1|08EE004|PASS|KEEP\n2|08EE001|LEVEL_ONLY|REJECT\n3|08EE005|FLOW_GAP|REJECT\n4|08EE003|FLOW_GAP|REJECT", "metadata": {"State-Gated Retrieval": ["First identify mainstem candidate stations on the Historical Results list that appear to cover 1948/1950 and carry a Flow label.", "A station qualifies only if its availability page shows the target years in the Flow section and all of April-July marked C.", "The final output must report target_window_status and final_verdict for each station in the same station set, ordered from downstream to upstream."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the Historical Results list yields four downstream-to-upstream mainstem candidates whose result rows span the target years and include Flow labels", "each candidate's availability page yields whether the target years are in the Flow table and whether April–July is complete", "the final ordered rows are produced only after all four stations are classified under the full Flow-table-and-Apr-Jul-C rule"], "control_dependency": ["list-level year spans and coarse Flow tags must be confirmed by explicit target-month checks on the availability page", "Hazelton, Smithers, and Houston candidate statuses must be fixed before the downstream-most KEEP row can be confirmed", "row order and KEEP/REJECT labels must be emitted only after the full availability-page rule is applied to all four stations"], "freeze": {"historical_window": "the target mainstem stations tested within the fixed historical daily Flow-availability windows for 1948 and 1950, with April–July complete-C coverage used as nearly stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["First identify mainstem candidate stations on the Historical Results list that appear to cover 1948/1950 and carry a Flow label.", "A station qualifies only if its availability page shows the target years in the Flow section and all of April-July marked C.", "The final output must report target_window_status and final_verdict for each station in the same station set, ordered from downstream to upstream."], "exclusion_conditions": ["Exclude stations deemed usable solely from list-level year spans and coarse Flow labels.", "Exclude results that mark any LEVEL_ONLY or FLOW_GAP station as KEEP.", "Exclude results that do not apply the full availability-page rule to all initially screened stations."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["rank", "station_number", "target_window_status", "final_verdict"], "rank": "emit 1 through 4 in downstream-to-upstream order within the validated station set", "station_number": "preserve the station number exactly", "target_window_status": "emit `PASS`, `LEVEL_ONLY`, or `FLOW_GAP` exactly", "final_verdict": "emit `KEEP` or `REJECT` exactly", "sorting_or_selection": "downstream to upstream within the validated Bulkley River mainstem station set", "stop_condition": ["all four initially screened stations are classified under the full availability-page rule", "the ordered table is emitted only after the downstream-to-upstream sequence is fixed"]}}}
{"task_id": "wateroffice_003-g", "domain": "WATER_OFFICE", "autonomy_type": "ordered table", "oracle_output_cardinality": 5, "instruction": "I am compiling historical snowmelt daily flow data for the Thompson River from Chase to Spences Bridge. I need to identify which mainstem stations along this reach actually had daily flow records for both 1958 and 1971 (May–July). For each mainstem station that appears in the station list as covering 1958 and 1971 and having flow information, verify it station by station in downstream order from Chase to Spences Bridge. Output a table with columns: rank|station_number|station_name|1958_MayJul|1971_MayJul|final_verdict. If a station has only water level records (no daily flow) for a target year, mark it LEVEL_ONLY. If daily flow records are missing or any of the May–July months is incomplete, mark it FLOW_GAP. Only if both years meet the criteria should final_verdict be KEEP; otherwise mark it REJECT.", "start_url": "https://wateroffice.ec.gc.ca/search/historical_e.html", "output_format": "Output a table with columns: rank|station_number|station_name|1958_MayJul|1971_MayJul|final_verdict.", "oracle_answer": "1|08LE031|SOUTH THOMPSON RIVER AT CHASE|PASS|PASS|KEEP\n2|08LE069|SOUTH THOMPSON RIVER AT MONTE CREEK|FLOW_GAP|LEVEL_ONLY|REJECT\n3|08LF023|THOMPSON RIVER AT KAMLOOPS|LEVEL_ONLY|LEVEL_ONLY|REJECT\n4|08LF033|THOMPSON RIVER NEAR SAVONA|FLOW_GAP|FLOW_GAP|REJECT\n5|08LF051|THOMPSON RIVER NEAR SPENCES BRIDGE|PASS|PASS|KEEP", "metadata": {"State-Gated Retrieval": ["All five candidate stations must be verified on their availability page: check whether the target year appears in the Flow table and whether May–July is entirely marked C.", "For each station, output a PASS, LEVEL_ONLY, or FLOW_GAP judgment for both target years, then label it KEEP or REJECT accordingly.", "The result order must follow the river-path sequence specified in the task."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["The river-path ordering yields five South Thompson/Thompson mainstem candidates.", "Each station's availability page indicates whether 1958 and 1971 are present in the Flow table and whether May–July is complete.", "Final rows are produced only after every candidate is classified under the stricter target-year and target-month Flow rule."], "control_dependency": ["Overall year coverage on the results list must be confirmed by the availability page inside the Flow table.", "Monte Creek, Savona, and Kamloops must be rejected by LEVEL_ONLY/FLOW_GAP evidence before the KEEP rows are final.", "The river-order sequence remains fixed, but station statuses must be assigned only after the availability-page rule is applied to all five stations."], "freeze": {"historical_window": "The ordered South Thompson/Thompson mainstem candidates are tested within the fixed historical Flow-availability windows for 1958 and 1971, with May–July complete-C coverage used as nearly stable reference information."}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["All five candidate stations must be verified on their availability page: check whether the target year appears in the Flow table and whether May–July is entirely marked C.", "For each station, output a PASS, LEVEL_ONLY, or FLOW_GAP judgment for both target years, then label it KEEP or REJECT accordingly.", "The result order must follow the river-path sequence specified in the task."], "exclusion_conditions": ["Exclude results that treat overall year coverage on the list page as sufficient evidence.", "Exclude results that fail to distinguish between Flow-table target-year and month completeness versus Level-only evidence.", "Exclude results that do not apply the same availability-page rule uniformly to all five stations."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["rank", "station_number", "station_name", "1958_MayJul", "1971_MayJul", "final_verdict"], "rank": "use the fixed Chase-to-Spences-Bridge mainstem order from the reference index, not the transient search-result order", "year_window_fields": "the two year-window fields emit only PASS, LEVEL_ONLY, or FLOW_GAP", "final_verdict": "emit `KEEP` or `REJECT` exactly"}}}
{"task_id": "wateroffice_004-g", "domain": "WATER_OFFICE", "autonomy_type": "ordered table", "oracle_output_cardinality": 13, "instruction": "I am compiling historical daily flow records for the main stem of the Columbia River from Fairmont Hot Springs to the International Boundary. I want to check, station by station, which Wateroffice gauges have complete daily flow records for May–October of 1955 and 1956. The candidate set is limited to main-stem Columbia River stations whose overall period of record spans both 1955 and 1956. Please evaluate each station in upstream-to-downstream order and output a table with columns: rank|station_number|station_name|1955_MayOct|1956_MayOct|final_verdict. If the target year has only water-level records and no daily flow records, mark it as LEVEL_ONLY. If daily flow records for May–October of that year are complete, mark it as PASS. If daily flow records are absent or any month from May to October has incomplete records, mark it as FLOW_GAP. Only when both 1955 and 1956 meet the complete daily flow condition should final_verdict be KEEP; otherwise, it should be REJECT.", "start_url": "https://wateroffice.ec.gc.ca/search/historical_e.html", "output_format": "Output a table with columns: rank|station_number|station_name|1955_MayOct|1956_MayOct|final_verdict.", "oracle_answer": "1|08NA045|COLUMBIA RIVER NEAR FAIRMONT HOT SPRINGS|PASS|PASS|KEEP\n2|08NA004|COLUMBIA RIVER AT ATHALMER|LEVEL_ONLY|LEVEL_ONLY|REJECT\n3|08NA052|COLUMBIA RIVER NEAR EDGEWATER|PASS|FLOW_GAP|REJECT\n4|08NA002|COLUMBIA RIVER AT NICHOLSON|PASS|PASS|KEEP\n5|08NB005|COLUMBIA RIVER AT DONALD|PASS|PASS|KEEP\n6|08NB006|COLUMBIA RIVER AT SURPRISE RAPIDS|PASS|PASS|KEEP\n7|08ND007|COLUMBIA RIVER ABOVE NAGLE CREEK|PASS|PASS|KEEP\n8|08ND011|COLUMBIA RIVER ABOVE STEAMBOAT RAPIDS|FLOW_GAP|PASS|REJECT\n9|08ND006|COLUMBIA RIVER AT TWELVE MILE FERRY|PASS|PASS|KEEP\n10|08NE002|COLUMBIA RIVER AT CASTLEGAR|LEVEL_ONLY|LEVEL_ONLY|REJECT\n11|08NE049|COLUMBIA RIVER AT BIRCHBANK|PASS|PASS|KEEP\n12|08NE003|COLUMBIA RIVER AT TRAIL|LEVEL_ONLY|LEVEL_ONLY|REJECT\n13|08NE058|COLUMBIA RIVER AT INTERNATIONAL BOUNDARY|PASS|PASS|KEEP", "metadata": {"State-Gated Retrieval": ["All candidate Columbia River stations must be checked on their availability page to confirm that the target years appear in the Flow table and that May–October is marked entirely as C.", "For each station, provide the KEEP/REJECT basis for both 1955 and 1956, distinguishing LEVEL_ONLY and FLOW_GAP cases from stations that only appear eligible from their overall year span.", "The final table must preserve the order and fields specified in the task."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the Columbia River candidate ordering yields fourteen stations spanning the target years", "each availability page yields whether 1955 and 1956 are present in the Flow table and whether May-October is complete", "final rows are produced only after every station is classified under the stricter Flow-table-and-May-Oct-C rule"], "control_dependency": ["list-level year spans must be confirmed by the availability page inside the Flow table", "LEVEL_ONLY and FLOW_GAP evidence must be applied to each station before the KEEP/REJECT table is finalized", "station-by-station verdicts and rejection reasons must be determined only after the full availability-page rule replaces the initial year-span screen"], "freeze": {"historical_window": "the ordered Columbia River candidates tested within the fixed historical Flow-availability windows for 1955 and 1956, with May-October complete-C coverage used as nearly stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["All candidate Columbia River stations must be checked on their availability page to confirm that the target years appear in the Flow table and that May–October is marked entirely as C.", "For each station, provide the KEEP/REJECT basis for both 1955 and 1956, distinguishing LEVEL_ONLY and FLOW_GAP cases from stations that only appear eligible from their overall year span.", "The final table must preserve the order and fields specified in the task."], "exclusion_conditions": ["Exclude results that retain a station solely because its overall period spans 1955/1956.", "Exclude results that fail to distinguish, station by station, between Flow-available, Level-only, and incomplete-month cases.", "Exclude results that do not apply the full availability-page rule uniformly to all 14 stations."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["rank", "station_number", "station_name", "1955_MayOct", "1956_MayOct", "final_verdict"], "rank": "use the fixed upstream-to-downstream Columbia mainstem order from the reference index, not the transient search-result order", "year_window_fields": "the two year-window fields emit only PASS, LEVEL_ONLY, or FLOW_GAP", "final_verdict": "emit `KEEP` or `REJECT` exactly"}}}
{"task_id": "wateroffice_005-g", "domain": "WATER_OFFICE", "autonomy_type": "ordered table", "oracle_output_cardinality": 3, "instruction": "I am organizing historical daily-flow records for the Peace River reach from Dunvegan Bridge to Fifth Meridian and want a three-row diagnostic table for the furthest-downstream reusable mainstem station, the first downstream failing station after it, and the provisional terminal station that an initial screen would misclassify as the downstream winner. If the target window satisfies the complete daily-flow condition, mark it as PASS. If the target year has only level records, mark it as LEVEL_ONLY. If daily-flow records exist but any target month from May through September is incomplete, mark it as FLOW_GAP. If the target year has neither daily-flow nor level records, mark it as NO_TARGET_RECORD. Output three rows with columns rank|station_number|diagnosis_role|target_window_status|provisional_count|final_count.", "start_url": "https://wateroffice.ec.gc.ca/search/historical_e.html", "output_format": "Output a three-row diagnostic table with columns rank|station_number|diagnosis_role|target_window_status|provisional_count|final_count.", "oracle_answer": "1|07HA001|FINAL_STATION|PASS|5|2\n2|07HD001|FIRST_BREAK_AFTER_FINAL|FLOW_GAP|5|2\n3|07KA002|PROVISIONAL_STATION|FLOW_GAP|5|2", "metadata": {"State-Gated Retrieval": ["Screen only among Peace River mainstem stations between Dunvegan Bridge and Fifth Meridian.", "A genuinely reusable station must satisfy the availability-page rule that both 1966 and 1967 appear in the Flow table and that every month from May through September is complete.", "The final output must jointly report FINAL_STATION, FIRST_BREAK_AFTER_FINAL, and PROVISIONAL_STATION, while repeating provisional_count and final_count in every row."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the Peace River mainstem chain between Dunvegan Bridge and Fifth Meridian yields the initial station list and the downstream breakpoint positions that need revalidation", "each availability page yields whether 1966 and 1967 are present in the Flow or Level tables and whether May-September is complete", "final rows are produced only after the final station, the first downstream failure, and the provisional downstream winner are all fixed under the same availability-page validation rule"], "control_dependency": ["station-level target-year Flow evidence must be validated by explicit May-September completeness checks on the availability page", "the final downstream reusable station and the first downstream failure must be determined under the same validation rule rather than from the initial station screen", "the provisional downstream winner is kept in the output only for reference after its FLOW_GAP status is confirmed on the availability page"], "freeze": {"historical_window": "the Peace River mainstem stations between Dunvegan Bridge and Fifth Meridian, evaluated within the fixed historical 1966 and 1967 Flow-availability windows, with May-September complete-C coverage used as nearly stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Screen only among Peace River mainstem stations between Dunvegan Bridge and Fifth Meridian.", "A genuinely reusable station must satisfy the availability-page rule that both 1966 and 1967 appear in the Flow table and that every month from May through September is complete.", "The final output must jointly report FINAL_STATION, FIRST_BREAK_AFTER_FINAL, and PROVISIONAL_STATION, while repeating provisional_count and final_count in every row."], "exclusion_conditions": ["Exclude results that treat a station as usable simply because the target years have Flow records.", "Exclude results that do not verify completeness month by month from May through September.", "Exclude results that mark a station such as 07KA002 as FINAL_STATION when availability-page evidence shows FLOW_GAP."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["rank", "station_number", "diagnosis_role", "target_window_status", "provisional_count", "final_count"], "rank": "emit the fixed order 1, 2, 3 for FINAL_STATION, FIRST_BREAK_AFTER_FINAL, and PROVISIONAL_STATION", "station_number": "use uppercase WSC station codes exactly as the Reference Index and availability pages show", "diagnosis_role": ["FINAL_STATION", "FIRST_BREAK_AFTER_FINAL", "PROVISIONAL_STATION"], "target_window_status": ["PASS", "LEVEL_ONLY", "FLOW_GAP", "NO_TARGET_RECORD"], "count_fields": "repeat the global provisional_count and final_count in every row as plain integers", "sorting_or_selection": "emit rows in the fixed diagnostic order FINAL_STATION, FIRST_BREAK_AFTER_FINAL, PROVISIONAL_STATION"}}}
{"task_id": "waterquality_001-g", "domain": "WATER_QUALITY_PORTAL", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I am compiling historical monitoring data for the middle Rio Grande and need to identify two monitoring locations suitable as sentinel sites for main-channel wastewater impacts. From the monitoring locations downstream of RIO GRANDE AT EMBUDO, NM that are confirmed to be on the Rio Grande mainstem, select two qualifying stations: the upstream station must be downstream of the Town of Bernalillo wastewater treatment plant outfall; the downstream station must be downstream of both the Town of Bernalillo and Rio Rancho No. 2 wastewater treatment plant outfalls. Both stations must simultaneously meet the following conditions: the station itself is not an outfall, facility, well, canal, or drainage ditch; the station type is labeled as River/Stream; records for Microbiological, Nutrient, Organics, Other, and Physical are all complete for 2015—2020; and the start years for Microbiological and Physical records are both earlier than 2010. If multiple mainstem stations in a given segment meet the conditions, select the one farthest upstream along the main channel. Output a single line in upstream-to-downstream order in the format <mainstem station ID>|<Microbiological start year>|<Physical start year>, with the two segments separated by a comma.", "start_url": "https://www.waterqualitydata.us/", "output_format": "Output a single line in upstream-to-downstream order in the format <mainstem station ID>|<Microbiological start year>|<Physical start year>, with the two segments separated by a comma.", "oracle_answer": "SANDIAWQ_WQX-RG003|2004|2004,SANDIAWQ_WQX-RG008|2004|2004", "metadata": {"State-Gated Retrieval": ["Select two target stations only from Rio Grande mainstem monitoring stations downstream of RIO GRANDE AT EMBUDO, NM, and require both to be past the specified outfall locations.", "The station detail page must explicitly indicate River/Stream, and records for Microbiological, Nutrient, Organics, Other, and Physical must fully cover 2015—2020, with start years for Microbiological and Physical both earlier than 2010.", "Output the two stations in upstream-to-downstream order."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["Provider/list views yield the Rio Grande mainstem candidates downstream of the Bernalillo and Rio Rancho No. 2 discharge points.", "Detail pages yield flow position, location type, characteristic-group period-of-record coverage for 2015—2020, and the Microbiological/Physical baseline years.", "Final rows are produced only after the corrected downstream sentry stations are fixed and ordered from upstream to downstream."], "control_dependency": ["Provider-list stream tags are only provisional and must be confirmed on the detail page by both mainstem placement and location type.", "Target-window presence is insufficient until full 2015—2020 coverage is confirmed on the detail page for all four groups.", "Once RG002/RG007 fail the full detail-page checks, the workflow must continue downstream from the discharge-point starting position to RG003/RG008."], "freeze": {"historical_window": "Rio Grande mainstem stations downstream of the Town of Bernalillo and Rio Rancho No. 2 WWTP outfalls, validated against 2015—2020 coverage for Microbiological, Nutrient, Organics, Other, and Physical."}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Select two target stations only from Rio Grande mainstem monitoring stations downstream of RIO GRANDE AT EMBUDO, NM, and require both to be past the specified outfall locations.", "The station detail page must explicitly indicate River/Stream, and records for Microbiological, Nutrient, Organics, Other, and Physical must fully cover 2015—2020, with start years for Microbiological and Physical both earlier than 2010.", "Output the two stations in upstream-to-downstream order."], "exclusion_conditions": ["Exclude stations that have a shallow (Stream) tag on the provider list or records near the target window, but whose detail page does not satisfy flow position, location type, or full period-of-record conditions.", "Exclude locations that are outfalls, facilities, wells, canals, or drainage ditches.", "Exclude candidates such as RG002 and RG007 that are not qualified sentinel stations downstream of the corresponding outfalls."], "normalization": {"record_separator": ",", "field_separator": "|", "schema_per_record": ["MonitoringLocationIdentifier", "Microbiological_start_year", "Physical_start_year"], "ordering": "emit exactly two records from upstream to downstream", "station_id": "use uppercase MonitoringLocationIdentifier", "year_fields": "emit four-digit years as plain integers", "selection": {"record_1": "the first qualifying mainstem station below the Bernalillo outfall", "record_2": "the first qualifying mainstem station below both the Bernalillo and Rio Rancho No. 2 outfalls"}, "stop_condition": ["the Bernalillo branch stops only after its first qualifying mainstem station is fixed", "the Rio Rancho No. 2 branch stops only after its first qualifying mainstem station is fixed", "the two segments are emitted as one comma-joined line in upstream-to-downstream order"]}}}
{"task_id": "waterquality_002-g", "domain": "WATER_QUALITY_PORTAL", "autonomy_type": "ordered table", "oracle_output_cardinality": 2, "instruction": "I am analyzing trends in the lower Rio Grande from 2001 to 2010. From monitoring stations located below RIO GRANDE BELOW ANZALDUAS and still within HUC 13090002, select one representative station for Hidalgo County and one for Cameron County. Each station must satisfy all of the following conditions: located below RIO GRANDE BELOW ANZALDUAS and still within HUC 13090002; the monitoring location type on the station detail page must be River/Stream; and records for all four characteristic groups—Microbiological, Nutrient, Organics, Other, and Physical—must cover the period 2001–2010. If multiple stations in the same county qualify, select the one with the earliest year in which all four groups are simultaneously present. Output a single line with the Hidalgo segment first and the Cameron segment second, using the format <County>:<MonitoringLocationIdentifier>|<earliest year all four groups present>|<Microbiological start year>|<Nutrient start year>|<Organics, Other start year>|<Physical start year>, with the two segments separated by a comma.", "start_url": "https://www.waterqualitydata.us/", "output_format": "Output a single line with the Hidalgo segment first and the Cameron segment second, using the format <County>:<MonitoringLocationIdentifier>|<earliest year all four groups present>|<Microbiological start year>|<Nutrient start year>|<Organics, Other start year>|<Physical start year>, with the two segments separated by a comma.", "oracle_answer": "Hidalgo:TCEQMAIN-15808|1998|1998|1998|1998|1998,Cameron:TCEQMAIN-13177|1995|1981|1981|1995|1981", "metadata": {"State-Gated Retrieval": ["Select one representative station for Hidalgo County and one for Cameron County only from monitoring stations located below RIO GRANDE BELOW ANZALDUAS and still within HUC 13090002.", "Candidate detail pages must show River/Stream as the monitoring location type, and records for all four characteristic groups—Microbiological, Nutrient, Organics, Other, and Physical—must cover 2001–2010.", "If multiple stations in the same county qualify, select the one with the earliest year in which all four groups are simultaneously present."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the fixed downstream starting point and county split yield Hidalgo and Cameron candidate mainstem stations within HUC 13090002", "detail pages yield River/Stream validity, the four characteristic-group start years, and the earliest year when all four groups are simultaneously present for 2001-2010 coverage", "final rows are produced only after the county-specific searches are rerun from the fixed downstream starting point and then ranked by common-start year"], "control_dependency": ["the Hidalgo search must preserve the \"below TCEQMAIN-13664\" downstream starting point rather than returning to the anchor station itself", "a station is only provisional until all four characteristic groups are confirmed for the same MonitoringLocationIdentifier and tidal/intake conflicts are excluded on the detail page", "county winners are valid only after late-start and non-river candidates are removed and the common-start-year tie-break is recomputed"], "freeze": {"historical_window": "stations below RIO GRANDE BELOW ANZALDUAS within HUC 13090002, split by Hidalgo and Cameron County, with Microbiological, Nutrient, Organics, Other, and Physical coverage over 2001-2010"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Select one representative station for Hidalgo County and one for Cameron County only from monitoring stations located below RIO GRANDE BELOW ANZALDUAS and still within HUC 13090002.", "Candidate detail pages must show River/Stream as the monitoring location type, and records for all four characteristic groups—Microbiological, Nutrient, Organics, Other, and Physical—must cover 2001–2010.", "If multiple stations in the same county qualify, select the one with the earliest year in which all four groups are simultaneously present."], "exclusion_conditions": ["Exclude results where the Hidalgo search loses the downstream starting point and brings the starting station itself back into the candidate set.", "Exclude results that treat a single multi-select of the four characteristic groups as sufficient without verifying that all four belong to the same station ID.", "Exclude stations with tidal, intake, late-start, or other conflicts clearly indicated on the detail page."], "normalization": {"record_separator": ",", "field_separator": "|", "schema_per_record": ["County:MonitoringLocationIdentifier", "common_start_year", "Microbiological_start_year", "Nutrient_start_year", "Organics_Other_start_year", "Physical_start_year"], "ordering": "emit exactly two records, with Hidalgo first and Cameron second", "county_prefix": "use the county label in title case, followed by : and the uppercase MonitoringLocationIdentifier", "year_fields": "emit four-digit years as plain integers", "common_start_year": "the earliest year in which all four required groups are simultaneously present for the chosen county station", "stop_condition": ["each county branch is rerun until one River/Stream station survives all four group checks", "the county winner is fixed only after late-start and non-river candidates are excluded on the detail page", "the two county segments are emitted on one line in the fixed Hidalgo-then-Cameron order"]}}}
{"task_id": "waterquality_003-g", "domain": "WATER_QUALITY_PORTAL", "autonomy_type": "ordered table", "oracle_output_cardinality": 3, "instruction": "I am compiling 2012–2018 water-quality reference data for the Brazos River mainstem downstream of Brazos Rv at Waco, TX. For each of the three specified HUCs that include Waco (12060202, 12070101, 12070104), I need to identify one TCEQ monitoring station suitable as a long-term reference for 2012–2018 along the Brazos River mainstem downstream of Brazos Rv at Waco, TX. Candidate stations must satisfy all of the following: be located within the corresponding HUC; have a monitoring-location type of River/Stream on the station detail page; have a station name that still corresponds to the Brazos River mainstem, not a tributary or other named waterbody; and have records covering 2012–2018 for all five characteristic groups—Inorganics, Major, Non-metals; Microbiological; Nutrient; Organics, Other; and Physical—meaning the station’s start year is no later than 2012 and its end year is no earlier than 2018. List the three results in upstream-to-downstream order, one per line, using the format: <HUC>|<MonitoringLocationIdentifier>|<earliest year when all five groups are jointly available>|<Inorganics, Major, Non-metals start year>|<Microbiological start year>|<Nutrient start year>|<Organics, Other start year>|<Physical start year>.", "start_url": "https://www.waterqualitydata.us/", "output_format": "List the three results in upstream-to-downstream order, one per line, using the format: <HUC>|<MonitoringLocationIdentifier>|<earliest year when all five groups are jointly available>|<Inorganics, Major, Non-metals start year>|<Microbiological start year>|<Nutrient start year>|<Organics, Other start year>|<Physical start year>.", "oracle_answer": "12060202|TCEQMAIN-12037|2000|1991|2000|1991|1999|1991\n12070101|TCEQMAIN-12030|2006|1972|1972|1972|2006|1972\n12070104|TCEQMAIN-11848|1975|1974|1975|1974|1975|1974", "metadata": {"State-Gated Retrieval": ["The search must stay anchored downstream of Waco and look for Brazos mainstem monitoring stations within the three HUCs specified in the task.", "Each candidate station must be checked separately across the five characteristic groups and then matched back to the same station ID, after which the detail page must verify full 2012–2018 coverage and mainstem semantics.", "For each HUC, retain only the qualifying station with the earliest common start year, and output the start-year fields required by the task."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the fixed downstream starting point below Waco and the three target HUC searches yield the Brazos mainstem candidate stations", "detail pages yield true downstream position, mainstem-vs-tributary semantics, and five-group period-of-record coverage across 2012-2018", "final rows are produced only after the three HUC-specific searches are carried out, matched by station ID, and ranked by the earliest year when all five groups are jointly available"], "control_dependency": ["the workflow must stay anchored downstream of Waco rather than reintroducing upstream stations such as TCEQMAIN-12044", "name-based waterbody matches must be validated on the detail page by excluding tributaries such as LITTLE BRAZOS RIVER and confirming full-window coverage", "station ranking must be determined after false positives from partial-window records are removed"], "freeze": {"historical_window": "Brazos River mainstem stations downstream of the Waco anchor evaluated within the fixed historical 2012-2018 window, using the HUC 12060202 / 12070101 / 12070104 split and five characteristic groups as largely stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["The search must stay anchored downstream of Waco and look for Brazos mainstem monitoring stations within the three HUCs specified in the task.", "Each candidate station must be checked separately across the five characteristic groups and then matched back to the same station ID, after which the detail page must verify full 2012–2018 coverage and mainstem semantics.", "For each HUC, retain only the qualifying station with the earliest common start year, and output the start-year fields required by the task."], "exclusion_conditions": ["Exclude results that lose the downstream starting point below Waco and bring upstream stations back into the candidate set.", "Exclude results that mistakenly treat tributaries such as LITTLE BRAZOS RIVER as the Brazos mainstem.", "Exclude stations that have data near the query window but do not cover the entire 2012–2018 window."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["HUC", "MonitoringLocationIdentifier", "common_start_year", "Inorganics_start", "Microbiological_start", "Nutrient_start", "Organics_Other_start", "Physical_start"], "HUC": "preserve the HUC exactly as emitted in the source result", "MonitoringLocationIdentifier": "preserve the monitoring-location identifier exactly, for example `TCEQMAIN-12037`", "common_start_year": "emit the maximum of the five per-group start years", "selection_rule": "only keep mainstem river/stream sites whose five required groups all cover the target 2012-2018 window", "sorting_or_selection": "upstream to downstream along the mainstem"}}}
{"task_id": "waterquality_004-g", "domain": "WATER_QUALITY_PORTAL", "autonomy_type": "ordered table", "oracle_output_cardinality": 3, "instruction": "I am reviewing toxic-pollutant-related water quality monitoring data for the Trinity River mainstem from 1993 to 2001, with the reach near FM 85 as the upstream boundary and the reach near Interstate 10 as the downstream boundary. I need to select one TCEQ long-term reference station for each of the upper, middle, and lower segments along the river.\nCandidate monitoring locations must simultaneously meet the following conditions: the monitoring location type in the station record is labeled River/Stream; the station name corresponds to the Trinity River mainstem, not a tributary, other named waterbody, or distributary channel; records for all six characteristic groups—Inorganics, Major, Non-metals; Microbiological; Nutrient; Organics, Other; Organics, PCBs; and Physical—must cover the period 1993–2001, specifically with a start year no later than 1993 and an end year no earlier than 2001. Output three rows in order from upstream to downstream along the river, each row formatted as <HUC>|<MonitoringLocationIdentifier>|<earliest year with all six groups>|<Inorganics, Major, Non-metals start year>|<Microbiological start year>|<Nutrient start year>|<Organics, Other start year>|<Organics, PCBs start year>|<Physical start year>.", "start_url": "https://www.waterqualitydata.us/", "output_format": "Output three rows in order from upstream to downstream along the river, each row formatted as <HUC>|<MonitoringLocationIdentifier>|<earliest year with all six groups>|<Inorganics, Major, Non-metals start year>|<Microbiological start year>|<Nutrient start year>|<Organics, Other start year>|<Organics, PCBs start year>|<Physical start year>.", "oracle_answer": "12030105|TCEQMAIN-10924|1983|1982|1983|1982|1983|1983|1982\n12030202|TCEQMAIN-10897|1993|1969|1972|1972|1993|1974|1969\n12030203|TCEQMAIN-10892|1976|1969|1971|1972|1974|1976|1969", "metadata": {"State-Gated Retrieval": ["Only retain candidate stations that satisfy the Trinity River mainstem semantics and require all six characteristic groups to cover 1993–2001.", "The six groups must be searched separately and then matched back to the same station ID, after which the detail page must verify full time-window coverage and completeness of the legacy toxics groups.", "For each target segment, retain only the station that meets all group requirements and has the best common start year."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["Trinity River mainstem candidate stations are generated from the target HUC searches and the six characteristic-group searches", "detail pages yield River/Stream validity, 1993–2001 full-window coverage, and the group-specific start years needed for ranking", "final rows are produced only after all six groups have been matched by station ID and detail-page validation removes partial-group and non-mainstem false positives"], "control_dependency": ["each of the six characteristic groups must be evaluated separately and confirmed against the same station ID before a multi-select result can be retained", "station names and River/Stream tags are insufficient until the detail page confirms Trinity mainstem semantics and 1993–2001 full-window coverage", "candidate rejection and final ranking must be determined after applying the PCBs / Organics, Other gap checks"], "freeze": {"historical_window": "Trinity River mainstem stations validated within the fixed historical 1993–2001 window, using six characteristic groups and ranking by earliest joint availability plus group start years as largely stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Only retain candidate stations that satisfy the Trinity River mainstem semantics and require all six characteristic groups to cover 1993–2001.", "The six groups must be searched separately and then matched back to the same station ID, after which the detail page must verify full time-window coverage and completeness of the legacy toxics groups.", "For each target segment, retain only the station that meets all group requirements and has the best common start year."], "exclusion_conditions": ["Exclude results that mistakenly treat a single multi-select of characteristic groups as sufficient to satisfy the requirement.", "Exclude results that only check station names and River/Stream tags without verifying 1993–2001 full-window coverage.", "Exclude stations that ignore missing PCBs or Organics, Other groups."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["HUC", "MonitoringLocationIdentifier", "common_start_year", "Inorganics_start", "Microbiological_start", "Nutrient_start", "Organics_Other_start", "Organics_PCBs_start", "Physical_start"], "HUC": "preserve the HUC exactly as emitted in the source result", "MonitoringLocationIdentifier": "preserve the monitoring-location identifier exactly, for example `TCEQMAIN-10924`", "common_start_year": "emit the maximum of the six per-group start years", "selection_rule": "only keep mainstem river/stream sites whose six required groups all cover the target window", "sorting_or_selection": "upstream to downstream along the mainstem"}}}
{"task_id": "waterquality_005-g", "domain": "WATER_QUALITY_PORTAL", "autonomy_type": "ordered table", "oracle_output_cardinality": 9, "instruction": "I am organizing integrated water-quality records for the lower Neches River during 1978-1984 and want a table of the nine Characteristic Group start years for the nearest qualifying TCEQ reference station downstream of NECHES RIVER AT CR 4915 that still represents the inland Neches River mainstem. Output a 9-row table with columns rank|MonitoringLocationIdentifier|HUC|County|characteristic_group|group_start_year|common_start_year, using the fixed nine-group order specified in the task.", "start_url": "https://www.waterqualitydata.us/provider/STORET/TCEQMAIN/TCEQMAIN-10598/", "output_format": "Output a 9-row table with columns rank|MonitoringLocationIdentifier|HUC|County|characteristic_group|group_start_year|common_start_year.", "oracle_answer": "1|TCEQMAIN-10580|12020003|Hardin County|Biological, Algae, Phytoplankton|1972|1977\n2|TCEQMAIN-10580|12020003|Hardin County|Inorganics, Major, Non-metals|1971|1977\n3|TCEQMAIN-10580|12020003|Hardin County|Inorganics, Minor, Metals|1974|1977\n4|TCEQMAIN-10580|12020003|Hardin County|Microbiological|1972|1977\n5|TCEQMAIN-10580|12020003|Hardin County|Nutrient|1972|1977\n6|TCEQMAIN-10580|12020003|Hardin County|Organics, Other|1974|1977\n7|TCEQMAIN-10580|12020003|Hardin County|Organics, PCBs|1974|1977\n8|TCEQMAIN-10580|12020003|Hardin County|Organics, Pesticide|1977|1977\n9|TCEQMAIN-10580|12020003|Hardin County|Physical|1971|1977", "metadata": {"State-Gated Retrieval": ["Search only among TCEQ monitoring stations downstream of NECHES RIVER AT CR 4915 that still represent the inland Neches River mainstem.", "A candidate station must cover 1978-1984 in all nine Characteristic Groups and, on the detail page, must still qualify as River/Stream, outside lake or reservoir waters, and outside the Sabine-Neches outlet boundary reach.", "The final output must be centered on the single winning station and list group_start_year plus common_start_year row by row in the fixed nine-group order."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["the fixed downstream starting point at NECHES RIVER AT CR 4915 yields candidate inland mainstem stations across successive downstream HUCs", "detail pages yield River/Stream validity, inland-vs-boundary semantics, and nine-group period-of-record coverage across 1978-1984", "final rows are produced only after one winning station survives the inland-mainstem checks and the requirement that all nine groups match the same station, after which its nine group start years are emitted in the required order"], "control_dependency": ["the nearest downstream mainstem station remains provisional until all nine characteristic groups have been checked separately and matched back to the same station ID", "the downstream search must continue to the next mainstem HUC when the first candidate fails the nine-group coverage requirement", "location semantics remain valid only after detail-page exclusion of lake/reservoir and Sabine-Neches boundary stations"], "freeze": {"historical_window": "Neches River inland mainstem stations downstream of NECHES RIVER AT CR 4915, validated within the fixed 1978-1984 window against nine characteristic groups as largely stable reference information"}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Search only among TCEQ monitoring stations downstream of NECHES RIVER AT CR 4915 that still represent the inland Neches River mainstem.", "A candidate station must cover 1978-1984 in all nine Characteristic Groups and, on the detail page, must still qualify as River/Stream, outside lake or reservoir waters, and outside the Sabine-Neches outlet boundary reach.", "The final output must be centered on the single winning station and list group_start_year plus common_start_year row by row in the fixed nine-group order."], "exclusion_conditions": ["Exclude results that retain TCEQMAIN-10591 merely because it appears to be the nearest downstream mainstem station.", "Exclude results that do not check the nine Characteristic Groups separately and verify that they all belong to the same station ID.", "Exclude stations located in lake or reservoir waters or in the Sabine-Neches outlet boundary reach."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "omit the header row; emit data rows only", "schema": ["rank", "MonitoringLocationIdentifier", "HUC", "County", "characteristic_group", "group_start_year", "common_start_year"], "rank": "emit 1 through 9 in the fixed characteristic-group order required by the task", "MonitoringLocationIdentifier": "use uppercase MonitoringLocationIdentifier", "HUC": "emit the HUC as an 8-digit string with no separators", "County": "use the county name exactly as shown on the detail page", "characteristic_group": "preserve the group label exactly as listed in the task", "year_fields": "emit four-digit years as plain integers", "common_start_year": "repeat the earliest year in which all nine required groups are simultaneously present for the chosen station", "sorting_or_selection": "emit rows in the fixed nine-group order stated in the task"}}}
{"task_id": "wonder_003-g", "domain": "CDC_WONDER", "autonomy_type": "ordered table", "oracle_output_cardinality": 8, "instruction": "For a public-health brief reviewing changes in U.S. home births in 2020, identify the months in which the home-birth rate among all mothers rose at least 25% compared with the same month in 2019. For those months, also provide the 2020 home-birth rates and relative changes for non-Hispanic Black and Hispanic mothers. Output a pipe-delimited table in calendar-month order with the fixed header row month|all_2019|all_2020|all_rel|black_2020|black_rel|hispanic_2020|hispanic_rel. Write all values as plain numbers without percent signs.", "start_url": "https://wonder.cdc.gov/natality-expanded-current.html", "output_format": "Output a pipe-delimited table in calendar-month order with the fixed header row month|all_2019|all_2020|all_rel|black_2020|black_rel|hispanic_2020|hispanic_rel. Write all values as plain numbers without percent signs.", "oracle_answer": "month|all_2019|all_2020|all_rel|black_2020|black_rel|hispanic_2020|hispanic_rel\nMay|1.03|1.49|45|0.74|54|0.55|57\nJune|1.04|1.31|26|0.66|20|0.48|30\nJuly|1.03|1.30|26|0.68|42|0.48|30\nAugust|1.02|1.29|26|0.69|50|0.48|26\nSeptember|1.01|1.31|30|0.73|43|0.54|54\nOctober|1.00|1.28|28|0.75|50|0.47|27\nNovember|0.99|1.30|31|0.73|43|0.55|67\nDecember|0.92|1.28|39|0.80|63|0.53|61", "metadata": {"State-Gated Retrieval": ["Compare the home-birth rates for all mothers in 2019 and 2020 by calendar month, retaining only months where the 2020 rate increased at least 25% relative to the same month in 2019.", "The home-birth numerator must use the full home category, not Home intended alone.", "For those qualifying months, also retrieve the 2020 home-birth rates for non-Hispanic Black and Hispanic mothers, along with their relative changes from the same month in 2019."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["Monthly natality queries yield the 2019 vs 2020 home-birth rates for all mothers and the candidate month set.", "Follow-up race/ethnicity queries yield the 2020 home-birth rates for non-Hispanic Black and Hispanic mothers, plus their same-month changes versus 2019.", "Final rows are produced only after qualifying months are determined from rate growth, and the resulting month set is reused for both subgroup backfill chains."], "control_dependency": ["Month selection must be based on relative rate increase rather than counts alone.", "The home-birth numerator must use Residence or an equivalent full-home measure rather than Home intended alone before downstream subgroup queries run.", "Subgroup columns must be generated from the resulting month set rather than appended to earlier rows."], "freeze": {"historical_window": "2019 and 2020 U.S. natality monthly home-birth rates for all mothers, non-Hispanic Black mothers, and Hispanic mothers."}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Compare the home-birth rates for all mothers in 2019 and 2020 by calendar month, retaining only months where the 2020 rate increased at least 25% relative to the same month in 2019.", "The home-birth numerator must use the full home category, not Home intended alone.", "For those qualifying months, also retrieve the 2020 home-birth rates for non-Hispanic Black and Hispanic mothers, along with their relative changes from the same month in 2019."], "exclusion_conditions": ["Exclude month-row sets that misinterpret \"rate increase of at least 25%\" as \"count increase of at least 25%.\"", "Exclude results that use only the Home intended category instead of the full home numerator.", "Exclude results that append subgroup columns based on an incorrect month set."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "required literal first row `month|all_2019|all_2020|all_rel|black_2020|black_rel|hispanic_2020|hispanic_rel`", "schema": ["month", "all_2019", "all_2020", "all_rel", "black_2020", "black_rel", "hispanic_2020", "hispanic_rel"], "month": "use the English month name and keep calendar order", "numeric_fields": "all numeric outputs are plain numbers without a percent sign", "relative_change_fields": "all_rel, black_rel, and hispanic_rel are emitted as integer percentages without the `%` character"}}}
{"task_id": "wonder_004-g", "domain": "CDC_WONDER", "autonomy_type": "ordered table", "oracle_output_cardinality": 3, "instruction": "From the 2021 U.S. infant mortality statistics by mother's race and Hispanic origin, identify all groups that are reported separately, have an infant mortality rate above the U.S. overall rate, and have a clearly identifiable leading cause of death. For each qualifying group, provide the infant mortality rate, neonatal mortality rate (under 28 days), postneonatal mortality rate (28 days and older), the group's leading cause of death, the cause-specific rate, and the rank of that cause among all infant causes. Output a multi-row pipe-delimited table sorted by infant mortality rate descending, using the format: group|infant_rate|neonatal_rate|postneonatal_rate|top_cause|cause_rate|us_rank.", "start_url": "https://wonder.cdc.gov/lbd-current-expanded.html", "output_format": "Output a multi-row pipe-delimited table sorted by infant mortality rate descending, using the format: group|infant_rate|neonatal_rate|postneonatal_rate|top_cause|cause_rate|us_rank.", "oracle_answer": "group|infant_rate|neonatal_rate|postneonatal_rate|top_cause|cause_rate|us_rank\nBlack non-Hispanic|10.55|6.35|4.19|Short gestation and low birth weight|196.4|2\nAmerican Indian or Alaska Native non-Hispanic|7.46|3.83|3.67|Congenital malformations|134.0|1\nPuerto Rican Hispanic|6.05|3.93|2.13|Short gestation and low birth weight|108.9|2", "metadata": {"State-Gated Retrieval": ["Retain only groups that are reported separately in the 2021 U.S. infant mortality statistics, have an infant mortality rate above the U.S. overall rate, and have a clearly identifiable leading cause of death.", "For each qualifying group, provide the infant rate, neonatal rate, postneonatal rate, top cause, cause rate, and the rank of that cause among all infant causes.", "The leading cause must be determined by a group-level leading-cause query and verified using notes/reliability flags."], "dependency_type": "Data + Control", "intra_chain": true, "inter_chain": true, "data_dependency": ["Infant-mortality queries yield group-level overall, neonatal, and postneonatal rates for reportable mother race/Hispanic-origin groups.", "Leading-cause queries and reliability notes yield each eligible group's top cause, cause-specific rate, and U.S. rank.", "Final rows are produced only after the eligible-group set is rebuilt using both overall-rate and cause-specific-identifiability criteria."], "control_dependency": ["Groups above the U.S. overall rate cannot be accepted until cause-specific reliability checks are run, which removes NHOPI.", "Leading cause must be queried at the group level rather than inherited from an all-infants cause table.", "The cause-rate multiplier must be defined as per 100,000 live births before downstream ranking/output."], "freeze": {"historical_window": "2021 U.S. infant mortality by mother race and Hispanic origin, including the group-level leading-cause tables and reliability notes."}, "answer_type": "multi-row ordered table"}, "rubric": {"inclusion_conditions": ["Retain only groups that are reported separately in the 2021 U.S. infant mortality statistics, have an infant mortality rate above the U.S. overall rate, and have a clearly identifiable leading cause of death.", "For each qualifying group, provide the infant rate, neonatal rate, postneonatal rate, top cause, cause rate, and the rank of that cause among all infant causes.", "The leading cause must be determined by a group-level leading-cause query and verified using notes/reliability flags."], "exclusion_conditions": ["Exclude groups that, despite having an infant mortality rate above the U.S. overall rate, cannot provide a clearly identifiable leading cause due to cause-specific reliability issues.", "Exclude results that inherit the top cause only from the all-infants cause table without performing a group-level leading-cause query.", "Exclude results where the cause rate uses an incorrect rate multiplier and is not read per 100,000 live births."], "normalization": {"field_separator": "|", "record_separator": "\n", "header_row": "required literal first row `group|infant_rate|neonatal_rate|postneonatal_rate|top_cause|cause_rate|us_rank`", "schema": ["group", "infant_rate", "neonatal_rate", "postneonatal_rate", "top_cause", "cause_rate", "us_rank"], "sorting_or_selection": "infant_rate descending", "rate_fields": "infant_rate, neonatal_rate, and postneonatal_rate are per 1,000 live births; cause_rate is per 100,000 live births, but all four are emitted as plain numbers without unit text", "top_cause": "emit the short cause label used in the oracle, not an expanded prose description", "us_rank": "emit the integer rank of that same cause in the overall infant leading-cause list"}}}