What this category's data looks like
This category's data typically consists of three types of materials: talent cultivation programs, course standards, and job competency analysis tables. Talent cultivation programs come from academic affairs offices and program leaders, are updated by academic year or cultivation cycle, and their fields include program name, program length, number of courses, graduation requirements, approval date, and signature; course standards come from course leaders, are updated by course version, and their fields include course name, credits, class hours, assessment method, and author; job competency analysis tables come from program development teams and enterprise research, are updated by job or research round, and their fields include job name, competency items, corresponding courses, and confirmation date. When the three types of materials enter this feature, each plays a different role: the talent cultivation program serves as the main record, while course standards and job competency analysis tables serve as comparison sources, used to locate discrepancies in the same object in terms of names, quantities, and dates. The lengths of the materials vary considerably, and comparison requires accommodating multiple file fragments, so the length of a single response and the range of context that can be accommodated become key constraints.
Configuration and implementation checks
This step is performed by application settings (response limit, etc.), which output the length of a single response 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 response; if set too low, the response will be truncated, and if set too high, consumption will increase.
| Object | Check | Verification action |
|---|---|---|
| Talent cultivation program | Check whether the program name, program length, number of courses, approval date, and signature are consistent with the course standards and job competency analysis table | Open the original program text, compare field by field with the other two materials, record the names and locations of inconsistent fields, and check whether the response length corresponding to maxResponse is sufficient to accommodate the list |
| Course standard | Check whether the course name, credits, class hours, assessment method, and author correspond to the course list in the program | Compare course by course, mark differences in name wording, differences in credits or class hours, unsynchronized versions, and missing signature items; check the tokens of quoted text separately using quoteMaxToken |
| Job competency analysis table | Check whether the job name, competency items, corresponding courses, and confirmation date connect with the program and course standards | Compare job by job, mark differences in the number of competency items, differences in corresponding course names, and missing confirmation dates; record the number of array entries separately, and do not compare by number of citations |
| Discrepancy list | Check whether each discrepancy states the object, field, values in the two materials, and source location | Review the list item by item, confirm that all three types of fields—names, quantities, and dates—have sources, and write back to the ledger after confirmation by the reviewer |
Three common errors
- Symptom: the same record has different course name wording in the program and the course standard. Cause: the two materials are maintained by different roles, names are not unified, and literal matching during comparison will miss discrepancies.
- Symptom: the course standard has been revised, but the program still references the old version's class hours and credits. Cause: versions are not synchronized, update rhythms are inconsistent, and the version number was not confirmed first during comparison.
- Symptom: the job competency analysis table lacks a confirmation date or author signature. Cause: no receipt was kept during material circulation, the signature column was not checked during archiving, and this item is empty in the discrepancy list.
How to verify the setup
Feature acceptance: ask a question that requires a long response, for example listing all discrepancy fields and their values in the two materials, check whether the end of the response is complete, and check whether the maxResponse and maxContext settings match that length.
Business test: use one real record to find fields inconsistent with the ledger, and record the object, field, values in the two materials, and source location.
- In application settings, check the response limit switch and slider position, and confirm that the limit matches the response length required by the discrepancy list.
- Compare usage from two runs, confirm the change in single-run usage after adjusting the limit, and check whether the end of longer responses is complete.