What this category's data looks like
The materials for annual progress and acceptance of research projects include annual progress reports, mid-term review materials, and acceptance reports. Annual progress reports are updated quarterly or semi-annually, recording stage objectives, task completion status, and budget execution; mid-term review materials are formed when the project is halfway through, containing indicator comparisons and expert opinions; acceptance reports summarize all achievements, test records, and signed receipts at project closure. After these materials enter the knowledge base retrieval and recall stage, each plays a different role: annual progress reports provide timeline fragments, mid-term review materials provide indicator verification fragments, and acceptance reports provide final conclusion fragments. Material fields involve names, quantities, dates, and signatures, and versions are updated as milestones progress. The retrieval service performs question optimization, multi-route recall, merging, reranking, filtering, and upper-limit trimming, returning candidate fragments and similarity scores for subsequent nodes or manual generation of handover-ready deployment status and operation 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 pattern, merging, reranking, filtering, and upper-limit trimming). Handover-ready deployment status and operation records are then completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Annual progress report | Fields include stage name, task quantity, start and end dates, and responsible person's signature, and must be aligned item by item with the mid-term review materials | In the search test, switch searchMode, test semantic, full-text, and hybrid modes separately with the same question, and record the number of candidates and scores |
| Mid-term review materials | Indicator names, completion quantities, review dates, and expert signatures must be consistent with the annual progress report, and version numbers must be synchronized | Turn usingReRank on and off, test once each, and compare the order and score changes of the top few candidates |
| Acceptance report | Achievement names, quantities, acceptance dates, and signed receipts must correspond to the mid-term review materials; missing receipts are considered incomplete | Adjust similarity and limit, observe the number of candidates and the tokens of quoted text, and record the array length and quotation budget separately |
| Deployment status and operation records | Version, connection, and task records must be consistent, and the handover form must be verifiable item by item | Check the handover form item by item once, see which record is missing, and record the number of candidates and scores |
Three common errors
- Phenomenon: The names and quantities of the same record do not match between the annual progress report and the mid-term review materials. Cause: The two materials are updated separately, and fields are not synchronized according to the same version.
- Phenomenon: The acceptance report lacks signatures or receipts, and the handover form cannot be closed. Cause: The signature process and material archiving are not synchronized, and receipts are not updated with the version.
- Phenomenon: The same question cannot be recalled when rephrased, or there are many candidates but the top-ranked ones are irrelevant. Cause: The retrieval mode does not match the material form, reranking is not enabled, or similarity filtering is too strict.
How to verify the setup
Acceptance of this feature: Test once each with synonym-rewritten questions, compare candidates and scores, then toggle reranking to check the ordering of the top few.
Business test: Check the handover form item by item once, see which record is missing, and verify whether version, connection, and task records are consistent.
- In the knowledge base search test, switch
searchMode, record the number of candidates and scores separately for the same question, and confirm that the multi-route merged results include both semantic and full-text sources. - Adjust
similarityandlimit, record the number of quotation array entries and the tokens of quoted text separately, and confirm that key fragments of the original text are still retained after trimming.