What this category's data looks like
This page addresses adult college entrance examination registration and admission management. The materials include adult college entrance examination registration forms, preferential scoring materials, and admission rosters. The registration form contains the candidate's basic information, application level and subject category, signature, and date; the preferential scoring materials contain the bonus item, certificate name, review opinion, and receipt; the admission roster contains the admission batch, major, planned number, and list of names. The materials enter the knowledge base in batches according to the three stages of registration, review, and admission, and the update cadence is synchronized with the batches. When entering the retrieval function, the registration form and bonus materials serve as sources of candidate chunks, and the admission roster serves as the basis for verification; the field names, quantities, and dates of the three must remain consistent across the two materials. If a signature or receipt is missing, supplement the record first before retrieval.
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 upper-limit trimming). The handover-ready deployment status and operation records are completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Adult college entrance examination registration form | Fields, dates, and signatures are consistent between the registration form and the admission roster | In the search test, switch searchMode, compare the candidate count and scores of the semantic, full-text, and hybrid options, and check item by item the names, quantities, dates, and signatures in the registration form and roster |
| Preferential scoring materials | The bonus item, certificate name, and review opinion match across the two materials | Compare the ranking of the top few results before and after enabling usingReRank, and verify whether the fields and receipts of the preferential scoring materials, registration form, and roster are consistent |
| Admission roster | The admission batch, major, and planned number are consistent between the roster and the registration form | Check whether the returned candidates include both semantic and full-text results, record the candidate count and scores, and verify the quantities and dates of the roster and registration form against the handover sheet |
| Retrieval parameters | After similarity filtering and citation upper-limit trimming, key original text chunks are still retained | Adjust similarity and limit, and record the candidate count and scores; check the tokens of the cited text separately, and record the array length separately |
Three common errors
- Symptom: The name, quantity, or date of the same candidate does not match between the registration form and the admission roster. Cause: When the two materials were entered, they were not checked against the same field definitions, and the versions were not synchronized.
- Symptom: The preferential scoring materials lack a signature or receipt, but retrieval still returns candidate chunks. Cause: The materials entered the knowledge base before completing the review workflow, and the handover sheet was not checked item by item.
- Symptom: The same question cannot be recalled when phrased differently, or there are many candidates but the ones ranked at the top are all irrelevant. Cause: The retrieval mode does not match the material form, and no secondary ranking was performed after hybrid retrieval.
How to verify the setup
This feature's acceptance: test once each with synonym-rewritten questions, compare candidates and scores, then toggle reranking on and off to inspect the ranking of the top few results.
Business test: check once item by item against the handover sheet, and see which record is missing.
- In the knowledge base search test, switch
searchMode, and compare the candidate count and scores of the semantic, full-text, and hybrid options. - Run once each with
usingReRankenabled and disabled, and record the changes in the order of the top few results; adjustsimilarityandlimit, and separately record the tokens of the cited text and the array length.