What this category's data looks like
This page addresses undergraduate program setup and dynamic adjustment materials, sourced from program setup application forms, talent training plans, expert review opinions, and receipt signature pages. The update cadence rolls with application batches and annual adjustment nodes, and the fields include program name, enrollment number, review date, responsible person's signature, version number, and attachment receipt. When the three types of materials enter knowledge base retrieval and recall, their roles differ: the application form provides the setup basis and quantity basis, the training plan provides course and program duration fields, and the expert review opinions provide conclusions and the signature chain. The retrieval service is only responsible for recalling candidate chunks and similarity scores; the anomaly list and review conclusions are generated by subsequent nodes or manually based on these. If material versions are not synchronized, the same record's name or date does not match, or signatures or receipts are missing, all of these leave traceable marks in the candidate chunks.
Configuration and implementation checks
This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate chunks and similarity scores (after mode, merge, rerank, filter, and limit trimming). The anomaly list and review conclusions are then completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Program setup application form | Name, quantity, date, and signature must be consistent with the training plan and review opinions | In the search test, switch searchMode, search with synonymous questions, record the number of candidates and scores, and check the application form fields against the source document |
| Talent training plan | Courses, program duration, and version number must be synchronized with the application form; a missing receipt means inconsistency | Compare the top few rankings before and after enabling usingReRank, and check whether the plan version and application form version are synchronized |
| Expert review opinions | Review date, conclusion, and signature must be complete; a missing signature or receipt is an anomaly | Check whether the candidates after similarity filtering contain the original review opinion text, and compare the date and signature page item by item |
| Anomaly record entry | Each anomaly must be traceable to the original record and material location | Record the number of candidates and scores, check the cited text tokens after trimming by limit, and record the array count separately |
Three common errors
- Phenomenon: The same program name is written differently in the application form and training plan, and the quantities do not match. Cause: The two versions of the materials were ingested separately, the versions were not synchronized, and retrieval hit their respective chunks.
- Phenomenon: The review opinion date differs from the application form date by several months, and the signature page is missing. Cause: The receipt was not imported together with the materials, and the candidate chunks only cover the main text.
- Phenomenon: There are many candidates, but the ones ranked first are all irrelevant, and the answer lacks key paragraphs. Cause: The retrieval mode does not match the material form, rerank is not enabled, or similarity filtering is too loose.
How to verify the setup
Acceptance for this feature: test once each with synonym-rewritten questions, compare candidates and scores, then toggle rerank on and off to inspect the ranking of the top few results.
Business test: create an anomaly record and see whether it can be traced back to the source material.
- In the knowledge base search test, switch retrieval modes, record the number of candidates and scores, and check the fields, dates, and signatures of the application form, training plan, and review opinions.
- Compare the order of the top few results before and after enabling rerank, and check whether the cited text retains key original passages after similarity filtering and citation limit trimming.