SGR-BENCH / data_en /constraint /census_003.json
Ninggggy's picture
Upload SGR-BENCH dataset files (without README variants)
8a1c7c6 verified
{
"task_id": "census_003",
"domain": "CENSUS_GOV",
"autonomy_type": "ordered table",
"oracle_output_cardinality": 4,
"instruction": "I am compiling a pre-pandemic county screening list for the California Department of Education to identify counties where, during annual school meal applications and parent data updates, schools cannot rely solely on English online portals and must still provide paper materials and in-person assistance. Use 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. First, select counties with a population aged 5–17 of at least 75,000. Then, retain counties with at least 2,000 individuals aged 5–17 who speak English \"not well\" or \"not at all\" and whose share of the 5–17 population exceeds the statewide share. Next, check these counties' digital access and income conditions: the share of households with no internet access must exceed the statewide share, and the median household income must be below the statewide median. Finally, among three offline support signals, retain counties where at least two exceed the statewide share: the share of households with no vehicle, the share of minors in households that received SSI, cash public assistance, Food Stamps, or SNAP in the past 12 months, and the share of renter-occupied housing units with more than one occupant per room. Output all qualifying counties in a single line, sorted alphabetically by county name. Each county should be formatted as County|SchoolAgePop|SchoolAgeLEP|LEPSharePct|NoInternetPct|NoVehiclePct|ChildBenefitSharePct|CrowdedRenterPct|MedianIncome|SupportPath, with counties separated by commas. SupportPath uses only the three labels BENEFITS, CROWD, and MOBILITY: the public assistance signal corresponds to BENEFITS, the crowded housing signal corresponds to CROWD, and the no-vehicle signal corresponds to MOBILITY. If multiple labels apply, join them with + in the fixed order BENEFITS+CROWD+MOBILITY.",
"start_url": "https://data.census.gov/advanced",
"output_format": "Output all qualifying counties in a single line, sorted alphabetically by county name. Each county should be formatted as County|SchoolAgePop|SchoolAgeLEP|LEPSharePct|NoInternetPct|NoVehiclePct|ChildBenefitSharePct|CrowdedRenterPct|MedianIncome|SupportPath, with counties separated by commas. SupportPath uses only the three labels BENEFITS, CROWD, and MOBILITY: the public assistance signal corresponds to BENEFITS, the crowded housing signal corresponds to CROWD, and the no-vehicle signal corresponds to MOBILITY. If multiple labels apply, join them with + in the fixed order BENEFITS+CROWD+MOBILITY.",
"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 of at least 2,000 with a share exceeding the statewide share, a no-internet share exceeding the statewide share, and a median household income below the statewide median.",
"SchoolAgeLEP must be the sum of the eight \"not well\" and \"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) exceeding 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 of at least 2,000 with a share exceeding the statewide share, a no-internet share exceeding the statewide share, and a median household income below the statewide median.",
"SchoolAgeLEP must be the sum of the eight \"not well\" and \"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) exceeding the statewide share."
],
"exclusion_conditions": [
"Exclude results that only take the two Spanish LEP rows and do not combine all eight target cells.",
"Exclude results that do not re-run the B28011, B19013, B08201, B09010, and B25014 branches.",
"Exclude counties whose final SupportPath does not have at least two signals exceeding 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."
]
}
}
}