What this category's data looks like
Vocational college skills competition organization involves three types of materials: competition regulations, registration forms, and training plans. Competition regulations are issued by the organizer and contain group, event, date, and signature/seal requirements; registration forms are submitted by participating institutions and contain contestant names, quantities, instructors, and acknowledgments; training plans are compiled by the coaching team and contain phase arrangements, venues, and signatures. Materials are updated according to the pre-competition, during-competition, and post-competition rhythm, with frequent version iterations. When entering the knowledge base retrieval service, regulations serve as the rule basis, registration forms serve as the record source, and training plans serve as process evidence; together they support the name, quantity, date, and signature verification of the same record, providing candidate fragments for discrepancy localization.
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 pattern, merge, re-rank, filter, and upper-limit trimming). The discrepancy list is then completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Competition regulations | The same event name, group, date, and signature/seal requirements are consistent between the regulations and the registration form | Switch searchMode in the search test, search using regulation field names, and check candidate fragments and similarity scores |
| Registration form | The same contestant names, quantities, instructors, and acknowledgments are consistent between the registration form and the training plan | Compare the order of the first few candidates before and after enabling re-rank, record the number of candidates and scores, and check signature and acknowledgment fields |
| Training plan | Phase dates, venues, and signatures are consistent between the training plan and the regulations | Test once each with synonymous questions, observe whether candidates cover key paragraphs, and check date and signature fields |
| Quoted content | After trimming, key fragments of the original text are still retained; the number of quotes and the text budget are recorded separately | Check the number and content completeness of returned quotes, verify the tokens of the quoted text, and do not compare the number of quotes and the budget as equal |
Three common errors
- Symptom: The name, quantity, and date of the same record do not match between the registration form and the training plan. Cause: The versions of the two materials are not synchronized, and retrieval candidates come from fragments of different versions.
- Symptom: The regulations and the registration form express the date of the same event inconsistently. Cause: The materials are updated at different rhythms, and old-version fragments are still recalled and not filtered.
- Symptom: The registration form lacks a signature or acknowledgment, and the training plan lacks a coach's signature. Cause: Material submission is incomplete, and retrieval candidates do not cover signature/seal fields.
How to verify the setup
Feature acceptance: test once each with synonymously rewritten questions, compare candidates and scores, then toggle re-rank to observe the ordering of the first few results, and switch searchMode to observe candidate changes, recording the number of candidates and scores.
Business test: use one real record to find fields inconsistent with the ledger, verify names, quantities, dates, and signatures, form a discrepancy list, submit it to the reviewer for confirmation, and write it back.
- In the search test, test once each with synonymous questions, compare candidate fragments and similarity scores, record the number of candidates, and adjust
similarityandlimitto observe the trimming effect. - Toggle re-rank to compare the ordering of the first few results, check the ordering changes before and after
usingReRank, confirm whether candidates cover key paragraphs, and then hand them over for manual generation of the discrepancy list.