What this category's data looks like
This page addresses the training institution enrollment consultation conversion scenario, and the materials come from three categories: enrollment brochures, consultation records, and trial class schedules. Enrollment brochures provide course names, number of class types, start dates, fee standards, and signature locations; consultation records carry prospective student names, communication dates, commitments, and receipts; trial class schedules record trial class times, instructors, attendance lists, and signature fields. Materials are updated by enrollment batch: brochures are revised as class types are adjusted, consultation records are appended item by item, and trial class schedules are arranged weekly. Field characteristics fall into four categories: names, quantities, dates, and signatures. Names include mixed use of full course names and abbreviations; quantities include planned headcount and actual enrolled headcount; dates include start dates and consultation dates; signatures include consultant signatures and student receipts. When the three material categories enter this feature, the brochure serves as the baseline version, consultation records serve as process evidence, and trial class schedules serve as implementation evidence, and the model completes field comparison and discrepancy localization within a single response.
Configuration and implementation checks
This part is executed by application settings (response limit, etc.), which output the length of a single response and the context range it can accommodate. The discrepancy list is then completed by the model, while entry and authorization are handled by the publishing service and file permissions.
| Object | Check | Verification action |
|---|---|---|
| Enrollment brochure | Course name, planned headcount, start date, and signature location align item by item with the consultation records | In application settings, check the response limit switch and slider position, confirm that a single response can accommodate the field list, and record the maxResponse value |
| Consultation records | Prospective student name, communication date, commitment quantity, and receipt signature align item by item with the trial class schedule | Compare the names and dates in the two materials, mark entries missing signatures or receipts, and separately record the number of reference array items and the quoteMaxToken budget |
| Trial class schedule | Trial class time, attendance count, instructor, and signature field align item by item with the brochure class types | Check the version number and update date, confirm old-version entries that have not been synchronized, and in long-response scenarios pay attention to whether maxContext is sufficient to accommodate multiple materials |
| Discrepancy list | The name, quantity, and date of the same object are consistent across the two materials | Group inconsistent fields by material source, hand them to the reviewer for confirmation, and write them back to the ledger |
Three common errors
- Phenomenon: the same student's name is written differently in the two materials. Cause: the brochure uses the full course name, while the consultation record uses an abbreviation, and name normalization has not been performed.
- Phenomenon: planned headcount and actual enrolled headcount do not match. Cause: the trial class schedule is updated weekly, the brochure version has not been synchronized, and quantity standards are mixed.
- Phenomenon: the receipt field is blank but the class has already been scheduled. Cause: missing signatures or receipts are not intercepted, and consultation records and trial class schedules are not cross-validated.
How to verify the setup
Feature acceptance: ask a question that requires a long response, check whether the ending is complete, and confirm that the response limit matches maxResponse.
Business test: use one real record to find fields inconsistent with the ledger, and verify the four items: name, quantity, date, and signature.
- In application settings, check the response limit switch and slider position, and confirm that the single response length matches the amount of material fields.
- Compare usage across two runs, confirm the change in single-run usage after the limit adjustment, and record the
maxContextandquoteMaxTokenvalues.