What this category's data looks like
This page focuses on the examination organization and violation handling category. Materials include examination manuals, examination room records, and violation fact records. Examination manuals are typically updated by academic year or examination batch, with fields including examination subjects, time nodes, organizational processes, and handling basis clauses; examination room records are generated per session, with fields including examination room number, expected and actual number of candidates, abnormal situations, invigilator signature, and date; violation fact records are generated per incident, with fields including candidate information, behavior description, evidence list, determination basis and handling opinion, and signatures of the parties and invigilators. The three types of materials play different roles when entering knowledge base retrieval: examination manuals provide cross-session general basis, examination room records provide session facts and headcount dates, and violation fact records provide case details and signature receipts. In terms of update rhythm, manuals are revised with system revisions, while records and fact records are archived after the examination ends. Version synchronization and signature completeness are key points for subsequent verification.
Configuration and implementation checks
This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate snippets and similarity scores (after pattern, merge, rerank, filter, and limit trimming). The basis entries that can be traced back to the original text are completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Examination manual | Clause names, examination batch dates, and current version numbers must match those in examination room records and violation fact records | In the search test, switch searchMode to semantic, full-text, or hybrid, and check whether the returned candidates contain the original manual clauses and version identifiers |
| Examination room record | Examination room number, expected and actual number of candidates, abnormal situation description, and invigilator signature date must be consistent with the violation fact record | Use the same session keywords for recall, check the number of candidate entries and similarity scores, and confirm that the headcount and date fields can be traced back to the original text of the record |
| Violation fact record | Candidate information, behavior description, determination basis, and signature receipt of the parties must be complete and correspond to manual clauses | Enable usingReRank to compare the ranking of the top entries, check whether the determination basis points to the current manual clauses and retains the signature receipt snippet |
| Retrieval returned candidates | Candidate snippets should retain key paragraphs of the original text; the number of cited entries and the citation content budget are recorded separately | Check the number of returned candidates and similarity, and observe whether the tokens of the trimmed citation text are complete according to limit |
Three common errors
- Phenomenon: The behavior name or date of the same violation incident does not match between the examination room record and the violation fact record. Cause: The two materials are entered separately without field alignment, and retrieval candidates hit different expressions.
- Phenomenon: The examination manual has been revised, but the examination room record still cites the old clause number. Cause: Versions are not synchronized; the old and current versions coexist in the knowledge base without effective dates marked.
- Phenomenon: The violation fact record lacks the signature or receipt of the parties, and the examination room record lacks the invigilator's signature. Cause: The archiving step does not validate signature fields; although retrieval candidates hit the fact description, they cannot be traced back to the complete original text.
How to verify the setup
Acceptance of this feature: Test once each with synonym-rewritten questions, compare candidates and similarity scores, then toggle usingReRank to see the ranking of the top entries; also switch searchMode to observe differences between semantic and full-text candidates.
Business test: Ask a cross-material basis question, see whether the answer provides traceable sources, and check that the number of cited entries and the tokens of the cited text are recorded separately.
- In the search test, fix the same question, switch
searchMode, and record the number of candidates andsimilarityscores, confirming that both semantic and full-text candidates can cover manual clauses and record fields. - Run once with
usingReRankenabled and once disabled, compare the order of the top entries; then check according tolimitwhether the trimmed citation text retains signatures, dates, and version identifiers.