What this category's data looks like
Graduate training process management involves three types of materials: training plans, proposal reports, and mid-term assessment forms. The sources are supervisors, colleges, and the graduate school, and the update cadence advances according to semester nodes and weekly records. Training plans include courses, credits, and supervisor signatures; proposal reports include title, date, review comments, and signature; mid-term assessment forms include progress, outcomes, assessment conclusions, and signature. When these three types of materials enter knowledge base retrieval and recall, each serves as an independent source fragment: training plans provide the basis for courses and credits, proposal reports provide topic selection and time nodes, and mid-term assessment forms provide stage progress and conclusions. Progress tables and statistical scope descriptions must be generated by subsequent nodes or manually based on these; retrieval only provides candidate fragments and similarity scores, and does not directly produce statistical tables.
Configuration and implementation checks
This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate fragments and similarity scores (after mode, merge, rerank, filter, and limit trimming). The "progress table and statistical scope description" is completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Training plan | Fields include course name, credits, date, and supervisor signature; the scope is written into the header | In the search test, switch searchMode, check whether candidates cover the course and credit fields, and record the number of candidates and scores |
| Proposal report | Fields include title, proposal date, review comments, and signature; the name and date must be consistent | Test once each with synonymous questions, see whether candidates hit the title and date, and compare the ranking of the top few items before and after enabling usingReRank |
| Mid-term assessment form | Fields include stage progress, outcomes, assessment conclusions, and signature; exceptions must be clearly stated | Check whether candidates contain both progress and conclusion fragments, and after adjusting similarity, record changes in the number of candidates and scores |
| Retrieval return | Outputs candidate fragments and similarity scores, not statistical tables or authorization results | Record the tokens of the cited text and the number of array items separately; limit only affects the citation upper bound, and does not verify source mismatches |
Three common errors
- Symptom: The name, quantity, and date of the same record do not match between the proposal report and the mid-term assessment form. Cause: The two materials are updated independently, without field-level alignment and version synchronization.
- Symptom: The training plan version is not synchronized, and old courses and credits are still recalled. Cause: The knowledge base was not re-imported according to semester nodes, and historical fragments are mixed into retrieval candidates.
- Symptom: The mid-term assessment form lacks a signature or receipt, but the statistical total is still adopted. Cause: The scope description did not list signatures and receipts as exceptions, and the original text was not reviewed item by item during verification.
How to verify the setup
Acceptance for this feature: test once each with synonymously rewritten questions, compare candidates and scores, then toggle rerank on and off to inspect the ranking of the top few items; check the number of candidates and scores corresponding to the three identifiers searchMode, usingReRank, and similarity.
Business test: use one week of records to verify the total, and see whether the scope and exceptions are clearly stated
- In the search test, check item by item whether candidate fragments contain course, date, and signature fields, and record the number of candidates and scores
- Record the tokens of the cited text and the number of array items separately; after adjusting
limit, only observe changes in the citation upper bound