What this category's data looks like
This page focuses on the scenario of general higher education institution admissions enrollment and filing, and organizes the field, date, and signature characteristics of three types of materials: enrollment plans, filing rules, and admission rosters. Enrollment plans include institution-major groups, planned numbers, subject selection requirements, and release batches; filing rules include filing ratios, tie-breaking sorting, batch time nodes, and version numbers; admission rosters include candidate numbers, filing scores, major preferences, and admission review signatures. Materials are updated by admissions year and batch. The names, quantities, and dates of the same record in two materials must be compared item by item. When versions are not synchronized or signature receipts are missing, the material must not enter external services. After the three types of materials enter knowledge base retrieval and recall, each plays a different role: enrollment plans and filing rules provide the basis for rules, while admission rosters provide the basis for results. Retrieval results serve only as candidate snippets and cited content for subsequent nodes or manual generation of externally available entry points and authorization scopes.
Configuration and implementation checks
This step is performed by the knowledge base retrieval service (retrieval node and search test), which outputs candidate snippets and similarity scores (after pattern, merge, rerank, filter, and upper-limit trimming). The externally available entry points and authorization scopes are then completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Enrollment plan | Whether the institution-major groups, planned numbers, subject selection requirements, and release dates are consistent with the filing rules | In the search test, switch searchMode to semantic, full-text, or hybrid, and check the plan fields and dates in the candidate snippets |
| Filing rules | Whether the filing ratio, tie-breaking sorting, batch time nodes, and version number correspond to the admission roster | Test once each with usingReRank enabled and disabled, and compare the ranking and similarity scores of the top candidate snippets |
| Admission roster | Whether the candidate number, filing score, major preferences, and admission review signature are complete and match the plans and rules | Access once with a restricted identity, check whether the visible scope matches the settings, and record the number of candidate snippets and scores |
| Authorization scope | Whether the visitor identity matches the visible scope, and whether unauthorized content is invisible | Test the same question once each with synonymous phrasings, see whether the candidate snippets and scores are stable, and check the number of cited items after similarity filtering |
Three common errors
- Phenomenon: The names, quantities, and dates of the same candidate in two materials do not match. Cause: The source material versions are not synchronized, and retrieval hits and associations were not checked first.
- Phenomenon: The answer lacks key paragraphs, and although there are many candidates, the ones ranked at the front are all irrelevant. Cause: Rerank is not enabled, or the similarity filter and citation upper-limit trimming are improperly set.
- Phenomenon: The visible scope of the external entry point exceeds authorization, and unauthorized content is recalled. Cause: The visitor identity and visible scope do not match, and materials missing signatures or receipts are not blocked.
How to verify the setup
This feature acceptance: Test once each with synonymously rewritten questions, compare the candidate snippets and scores, then toggle rerank on and off to observe the ranking of the top few; record the number of candidate snippets and scores, and when citation upper limits are involved, check the tokens of the cited text, recording array item counts separately.
Business test: Access once with a restricted identity, see whether the visible scope matches the settings; check whether the fields, dates, and signatures of the enrollment plans, filing rules, and admission rosters are consistent.
- In the knowledge base search test, switch
searchModeand compare the candidate snippets and similarity scores of the semantic, full-text, and hybrid modes. - Test once each with
usingReRankenabled and disabled, check the number of cited items and content completeness aftersimilarityfiltering, and verify whetherlimittrimming preserves the key snippets of the original text.