What this category's data looks like
Investigation and handling materials for academic misconduct reports usually comprise three categories: report materials, investigation reports, and expert appraisal opinions. Fields involve the reported person, the matter name, the quantity involved, the date of occurrence, signatures, and receipts. Material sources are mostly the acceptance registration, investigation and evidence collection, and expert deliberation stages. The update rhythm follows case progress with supplementary entries added in batches, and the same record often shows mismatched names, quantities, and dates across different materials. When entering the knowledge base retrieval and recall function, report materials assume the role of factual clues, investigation reports assume the role of process and conclusions, and expert appraisal opinions assume the role of professional judgment. Together the three constitute the retrieval corpus, for subsequent nodes or manual retrieval of fragments to generate a handover-ready deployment state and operation records.
Configuration and implementation checks
This link is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate fragments and similarity scores (after mode, merge, re-rank, filter, and upper-limit trimming). The handover-ready deployment state and operation records are completed subsequently by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Report materials | Fields, dates, and signatures are complete, and names and quantities match the investigation report | In the search test, switch searchMode to compare candidates, and verify the reported person, matter, quantity, date, and signature receipt |
| Investigation report | Version is in sync with the expert appraisal opinions, with no missing signatures or receipts | Verify whether the report version number is consistent with the appraisal opinion version, and check the signature field and receipt records |
| Expert appraisal opinions | Appraisal conclusions correspond to the fragments cited in the investigation report, and dates are traceable | Retrieve once each with synonymous questions, see whether candidate fragments fall within the original text of the appraisal opinions, and record the number of candidates and scores |
| Retrieval configuration | similarity, limit, and usingReRank match the material form | Check the number of returned candidates and scores, compare the order of the top few entries before and after enabling re-rank, and separately record the length of the citation array |
Three common errors
- Phenomenon: The name, quantity, and date of the same record do not match between the report materials and the investigation report. Cause: The two materials were entered in batches without field alignment, and the original text was not verified after retrieval hits.
- Phenomenon: The investigation report version and the expert appraisal opinion version are not in sync, and the handover form is missing one record. Cause: The update rhythms are not synchronized, and the old version remains in the knowledge base participating in recall.
- Phenomenon: The report materials lack signatures or receipts, and the responsible stage cannot be confirmed at handover. Cause: Acceptance and evidence collection records were not entered into the knowledge base together with the materials, and retrieval returns only body fragments.
How to verify the setup
This function's acceptance: test once each with synonymously rewritten questions, compare candidate fragments and similarity scores, then toggle re-rank on and off to observe changes in the ordering of the top few entries, and record the number of candidates, scores, and the length of the citation array.
Business test: check item by item against the handover form once, see which record is missing, and confirm that the version, connection, and task records are consistent.
- In the knowledge base search test, switch
searchMode, retrieve once each for multiple phrasings of the same question, and record the number of candidates and scores. - After verifying the
similarityandlimitsettings, enableusingReRankto compare the order of the top few entries, and separately record the length of the citation array and the tokens of the cited text.