What this category's data looks like
University laboratory safety grading and classification materials mainly consist of the Laboratory Grading and Classification Table, the Hazard Source List, and the Safety Responsibility Statement. They come from departmental submissions and school-reviewed archiving, and their update cadence adjusts with experimental projects, hazardous chemical inventory, and changes in responsible persons. When the three materials enter knowledge base retrieval, their roles differ: the grading and classification table provides core fields such as laboratory level, room number, and hazard category; the hazard source list provides specific substances, quantities, storage locations, and control measures; the safety responsibility statement provides the responsible person, signature date, and receipt status. Field characteristics concentrate in four categories: name, quantity, date, and signature. The same record across the three materials is prone to inconsistencies in name spelling, quantity basis, and date order. When versions are not synchronized, the old version is still recalled, and missing signatures or receipts break the responsibility chain. The knowledge base retrieval service splits fragments by material version, and during retrieval returns candidate fragments and similarity scores for subsequent nodes or manual verification of the version ledger and change records.
Configuration and implementation checks
This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate fragments and similarity scores (after mode, merge, rerank, filter, and limit trimming). The version ledger and change records are then completed by subsequent nodes or manually; the retrieval results themselves are not statistical tables, business documents, or authorization results.
| Object | Check | Verification action |
|---|---|---|
| Laboratory Grading and Classification Table | Check whether the laboratory name, room number, level, hazard category fields, and approval date are consistent with the current version | In the search test, switch searchMode to semantic, full-text, or hybrid, compare candidate fragments and similarity scores, and confirm that the grading and classification table fields are hit |
| Hazard Source List | Check whether the substance name, quantity, storage location, control measures, and date match the records in the grading and classification table | Turn usingReRank on or off, compare the order of the top candidate fragments, and check whether hazard source list fragments stably rank at the front |
| Safety Responsibility Statement | Check whether the responsible person, signature date, and receipt status are complete and synchronized with the current version | Adjust the similarity filter threshold, observe whether responsibility statement fragments are filtered out, and then record the number of candidate fragments and scores |
| Version Ledger and Change Records | Check that the current version is unique, old versions are traceable, references are synchronized, and change records cover the three materials | Use limit to control the reference cap, separately record the number of items in the reference array, and check whether the reference text tokens retain key fields |
Three common errors
- Phenomenon: The same laboratory's name, quantity, and date do not match between the grading and classification table and the hazard source list. Cause: The two materials were submitted by different personnel at different times, and no field-level cross-check was performed.
- Phenomenon: The safety responsibility statement lacks a signature or receipt, yet retrieval still recalls old-version fragments. Cause: Versions are not synchronized, the old version is not marked invalid, and retrieval returns candidates by similarity.
- Phenomenon: Changing the wording of a question makes it impossible to recall key passages, and although there are many candidates, the ones ranked at the front are irrelevant. Cause: The retrieval mode does not match the material form, and rerank and similarity filtering were not adjusted according to field characteristics.
How to verify the setup
Acceptance for this feature: test once each with synonym-rewritten questions, compare candidate fragments and similarity scores, then toggle rerank on and off to observe the ordering of the top few results, and check that the corresponding status values for searchMode, similarity, and usingReRank record only the number of candidate fragments and scores.
Business test: change one piece of content, see whether the version number and references are updated synchronously, and confirm that the current version is unique, old versions are traceable, and references are synchronized.
- In the knowledge base search test, retrieve the same question using semantic, full-text, and hybrid modes respectively, record the number of candidate fragments and similarity scores, and check whether the fields of the grading and classification table, hazard source list, and safety responsibility statement are hit.
- Compare the order of the top candidate fragments before and after enabling rerank, check the number of items in the reference array and the reference text tokens, and confirm that the version ledger and change records are then completed by subsequent nodes or manually.