What this category's data looks like
In the business of general higher-education institution admissions and enrollment filing, three types of materials—admission plans, filing rules, and admission rosters—are the core sources for basis queries. Admission plans usually state the institution-major group, planned number, selected examination subjects, and batch; filing rules state the filing ratio, tie-breaking ranking, withdrawal conditions, and time nodes; admission rosters state the candidate number, admitted major, signature, and receipt. Materials are updated by year and batch, and the currently effective version prevails. When entering the knowledge base retrieval and recall function, admission plans and filing rules mainly serve as candidate sources for rule clauses and numerical bases, while admission rosters serve as candidate sources for case records and signature receipts; the three types of materials jointly support the recall of cross-material basis entries, and the subsequent nodes or humans ultimately generate basis entries that can be traced back to the original text.
Configuration and implementation checks
This step is executed by the knowledge base retrieval service (retrieval node and search test), which outputs candidate snippets and similarity scores (after mode, merge, re-rank, filter, and upper-limit trimming). Basis entries that can be traced back to the original text are then completed by subsequent nodes or humans. The retrieval service itself does not produce statistical tables, business documents, or authorization results.
| Object | Check | Verification action |
|---|---|---|
| Admission plan | Whether the planned number, selected examination subjects, batch, and date match the currently effective version, and whether the institution-major group name is consistent | In the search test, search by institution-major group and batch, check the planned number and date in the candidate snippets, record the number of candidates and scores, and use searchMode to switch modes for comparison |
| Filing rules | Whether the filing ratio, tie-breaking ranking, withdrawal conditions, and time nodes come from the currently effective materials of this institution and can be traced back to the original text | Test once each with synonymous questions, check whether the candidate snippets cover the rule clauses and time nodes, and enable usingReRank to compare the ranking of the top few results |
| Admission roster | Whether the candidate number, admitted major, signature, and receipt fields are complete, and whether the name, quantity, and date of the same record are consistent across the two materials | Search for the same candidate record across materials, check the names and dates in the roster against the plan and rules, record the number of candidates and scores, and if necessary use similarity to filter low-scoring candidates |
| Version and receipt | Whether the material version is currently effective, and whether entries missing a signature or receipt are mistakenly recalled | Limit the search scope by year and batch, check the version identifier and signature/receipt fields in the candidate snippets, and confirm that key snippets of the original text are still retained after trimming |
Three common errors
- Symptom: The name, quantity, or date of the same record does not match between the admission plan and the admission roster. Cause: Material versions are not synchronized or entry standards are inconsistent, and after a retrieval hit, the original text was not checked by tracing back.
- Symptom: Recall of filing rule clauses is missing, and the answer lacks key paragraphs. Cause: The retrieval mode does not match the material form, or similarity filtering and citation upper-limit trimming are excessive.
- Symptom: Admission roster entries missing a signature or receipt are still treated as a basis. Cause: The candidate snippets do not cover the signature/receipt fields, and subsequent nodes or humans did not verify by tracing back to the original text.
How to verify the setup
Acceptance of this function: test once each with synonymously rewritten questions, compare the number of candidates and scores, then toggle re-ranking on and off to observe the ranking of the top few results.
Business test: ask a cross-material basis question and see whether the answer provides a traceable source.
- In the search test, switch
searchModeto compare the number of candidates and scores among semantic, full-text, and hybrid modes, and record whether the candidates include results from multiple paths at the same time. - Compare the ranking of the top few results before and after enabling
usingReRank, and usesimilarityto filter low-scoring candidates, confirming that the cited text still retains key snippets of the original text after trimming.