What this category's data looks like
This page focuses on the compulsory education dropout prevention and control scenario, with materials drawn from three categories: dropout risk ledgers, return-to-school persuasion records, and home visit records. The ledger is updated monthly and contains student name, risk level, responsible person, and date; the return-to-school persuasion record is filled in with each persuasion, and contains the subject name, number of persuasions, result, and signature; the home visit record is registered per visit, and contains the visited student, home visit date, problem description, and receipt. When the three categories of materials enter this function, the ledger serves as the baseline object, and the return-to-school persuasion and home visit records serve as comparison objects, each taking on the roles of field source and discrepancy localization. Field characteristics are concentrated in four categories: name, quantity, date, and signature. The name must be consistent with the student status; quantity refers to the number of persuasions or home visits; the date must correspond to the ledger node; the signature and receipt are used to confirm that the process is closed-loop. When material versions are not synchronized, the name, quantity, and date of the same record in the two materials are prone to mismatch, requiring field-by-field verification and the formation of a discrepancy list.
Configuration and implementation checks
This part is executed by application settings (response limit, etc.), which output the length of a single answer and the range of context that can be accommodated. The "discrepancy list" is then completed by the model. The response limit uses a slider to control the length of a single answer. If set too low, the answer will be truncated; if set too high, consumption will increase.
| Object | Check | Verification action |
|---|---|---|
| Dropout risk ledger | Whether the name, quantity, date, and signature are consistent with the return-to-school persuasion record | Compare name, risk level, number of persuasions, and date item by item, record discrepant fields and check the signature column; pay attention to the truncation impact of maxResponse on long list output |
| Return-to-school persuasion record | Whether the name, quantity, and date of the same subject match in the two materials | Using the ledger as the baseline, check the number of persuasions, persuasion result, and receipt date, and mark entries whose versions are not synchronized or whose signatures are missing |
| Home visit record | Whether the home visit date, problem description, and receipt correspond to the ledger nodes | Check the number of home visits and the ledger node dates, and verify the receipt signature; when outputting long lists, pay attention to the budget allocation of maxContext and quoteMaxToken |
| Discrepancy list | Whether field-level discrepancies can be located to specific materials and rows | Summarize the four categories of discrepancies: name, quantity, date, and signature, submit them to the reviewer for confirmation, and write them back |
Three common errors
- Phenomenon: The number of persuasions for the same record does not match between the ledger and the return-to-school persuasion record. Cause: Versions are not synchronized; after one side is updated, the other side is not written back, resulting in a discrepancy in the quantity field.
- Phenomenon: The date in the home visit record is inconsistent with the ledger node date. Cause: It was not filled in according to the ledger node at registration, or the date was entered incorrectly during supplementary entry, causing the date for the same subject to mismatch.
- Phenomenon: The return-to-school persuasion record lacks a signature or receipt. Cause: The process is not closed-loop; the signature and receipt are not synchronized with the record, so it cannot be confirmed during verification that the record has been processed.
How to verify the setup
This function's acceptance check: ask a question that requires a long answer, for example listing all discrepant fields, and see whether the end of the answer is complete, confirming that the response limit has not truncated it.
Business test: use one real record to find fields inconsistent with the ledger, and record the specific discrepancies in name, quantity, date, or signature.
- In the application settings, check the response limit switch and slider position, and confirm that they match the answer length required by the page.
- Compare usage across two runs, observe the change in single-run usage after adjusting the limit, and check whether the end of longer answers is complete.