What this category's data looks like
Adult college entrance examination registration and admission management involves three types of materials: registration forms, preferential bonus point materials, and admission rosters. They are respectively filled out by candidates, issued by review departments, and archived by admissions institutions, and are updated on a rolling basis according to registration seasons, review milestones, and admission batches. Registration forms contain name, ID number, application level, subject category, photo, and signature; preferential bonus point materials contain bonus item, certificate number, issue date, and handler seal; admission rosters contain candidate number, admitted major, admission date, and responsible person's signature. When these three types of materials enter knowledge base retrieval and recall, each plays a different role: registration forms provide the baseline for identity and application fields, bonus materials provide evidence of eligibility and dates, and admission rosters provide proof of results and signatures. Field characteristics manifest as three types of values—names, quantities, and dates—that need to be aligned across materials, while signatures and receipts constitute integrity constraints.
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 pattern, merge, rerank, filter, and limit trimming). The discrepancy list is then completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Registration form | Whether name, ID number, application level, subject category, and signature are consistent with the ledger | In the search test, use synonymous questions to retrieve registration form fields, record the number of candidates and scores, and verify signatures and receipts |
| Preferential bonus point materials | Whether bonus item, certificate number, issue date, and handler seal correspond to the registration form | Switch searchMode to hybrid, compare semantic and full-text candidates, and verify whether dates and quantities are aligned |
| Admission roster | Whether candidate number, admitted major, admission date, and responsible person's signature are consistent with the registration form | Compare the ranking of the top few results before and after enabling usingReRank, and verify name and date fields |
| Discrepancy list | Whether the name, quantity, and date of the same object are consistent across the two materials | Filter low-scoring candidates by similarity, record inconsistent fields one by one, and submit them to the reviewer for confirmation and write-back |
Three common errors
- Phenomenon: The same candidate's name is written differently in the registration form and the admission roster. Cause: Versions are not synchronized, the roster still uses old registration information, and retrieval candidates do not cover the latest fields.
- Phenomenon: The certificate number in the bonus materials does not match the quantity in the registration form. Cause: Materials are submitted in batches, date and quantity fields are not updated with the batches, and candidate chunks come from different versions.
- Phenomenon: The admission roster lacks the responsible person's signature or receipt. Cause: The signature page was not scanned into the knowledge base, retrieval only recalls the main text, and the integrity constraint is not reflected in candidate scores.
How to verify the setup
Acceptance of this feature: test once each with synonymously rewritten questions, compare candidates and scores, then toggle rerank to observe the ranking of the top few results; record the status of searchMode, similarity, and usingReRank, and separately record the number of citations and cited text tokens.
Business test: use one real record to find fields inconsistent with the ledger, and submit them to the reviewer for confirmation and write-back.
- In the knowledge base search test, switch
searchMode, use semantic, full-text, and hybrid recall for the same field respectively, and record differences in the number of candidates and scores. - Enable and disable
usingReRank, compare the order of the top few candidates, and after filtering bysimilarity, verify cited text tokens and the number of citations.