| { |
| "task_id": "consumerfinance_012", |
| "domain": "CFPB_COMPLAINT_DATABASE", |
| "autonomy_type": "ordered table", |
| "oracle_output_cardinality": 2, |
| "instruction": "我在做一份历史投诉簇的内部 outreach memo，想追踪一组比较特殊的投诉：在 CFPB Complaint Database 里，找出当前那个覆盖 credit reporting 和其他 personal consumer reports 的产品桶下、2018-01-01 到 2021-12-31 之间、标签是 Servicemember、有公开 narrative、并且文本里明确提到 medical bills 或 doctors 的投诉。请给我两条代表性投诉簇：第一条是这个 cohort 的峰值月份；第二条是这个 cohort 里最新、且投诉量仍至少达到峰值一半的月份。对每条投诉簇，请告诉我该月的总投诉量、该月投诉量最高的州及其投诉量、在这个州里该月投诉量最高的公司及其投诉量、以及这家公司在这个州和这个月里最早收到的那条投诉的编号、日期、Issue 和 Sub-issue。请按 Peak 在前、LatestHalfPeak 在后输出一行字符串：<Cluster>|<Month>|<MonthComplaints>|<TopState>|<StateComplaints>|<TopCompany>|<CompanyComplaints>|<EarliestComplaintID>|<EarliestDate>|<Issue>|<SubIssue>。", |
| "start_url": "https://www.consumerfinance.gov/data-research/consumer-complaints/", |
| "output_format": "请按 Peak 在前、LatestHalfPeak 在后输出一行字符串：<Cluster>|<Month>|<MonthComplaints>|<TopState>|<StateComplaints>|<TopCompany>|<CompanyComplaints>|<EarliestComplaintID>|<EarliestDate>|<Issue>|<SubIssue>。Peak 簇与 LatestHalfPeak 簇用英文逗号连接；若存在并列，月份按时间顺序判定，州按州缩写字母序升序取首位，公司按公司名称字母序升序取首位，最早投诉若同日并列则取 Complaint ID 最小者。", |
| "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": [ |
| "只看当前 product bucket 下 credit reporting / other personal consumer reports 相关产品中，2018-01-01 至 2021-12-31、标签为 Servicemember、带公开 narrative 且文本明确提到 medical bills 或 doctors 的投诉。", |
| "第一条 cluster 必须是该 cohort 的 peak month；第二条 cluster 必须是该 cohort 中最新且投诉量仍至少达到峰值一半的月份。", |
| "对每个 cluster，都必须在对应的 month + state 子集中重算 TopCompany，再定位该公司在该州该月最早收到的 complaint。" |
| ], |
| "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": [ |
| "只看当前 product bucket 下 credit reporting / other personal consumer reports 相关产品中，2018-01-01 至 2021-12-31、标签为 Servicemember、带公开 narrative 且文本明确提到 medical bills 或 doctors 的投诉。", |
| "第一条 cluster 必须是该 cohort 的 peak month；第二条 cluster 必须是该 cohort 中最新且投诉量仍至少达到峰值一半的月份。", |
| "对每个 cluster，都必须在对应的 month + state 子集中重算 TopCompany，再定位该公司在该州该月最早收到的 complaint。" |
| ], |
| "exclusion_conditions": [ |
| "排除把“最新有投诉的月份”直接当成第二条 cluster 的结果。", |
| "排除先在全 cohort 中找 overall top company 再投射到 month-state 子集的结果。", |
| "排除没有先用 PeakCount 推出 half-peak threshold 的结果。" |
| ], |
| "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" |
| } |
| } |
| } |
|
|
