What this category's data looks like
This category's data for the research project task book signing scenario includes project task books, subject task breakdown tables, budget books, and accompanying signature pages and receipts. The sources are research management departments, subject leaders, and finance and audit archives; the update cadence advances with signing rounds, version revisions, and supplementary signing receipts; and the field characteristics center on subject name, undertaking unit, leader, start and end dates, funding subjects, quantities, and signature fields. When the materials enter the knowledge base retrieval and recall function, the task book plays the role of the main record, the breakdown table plays the role of splitting subjects and task levels, the budget book plays the role of funding calibers and subject mapping, and the signature pages and receipts play the role of signing status and version synchronization. Together, the four types of materials constitute the source of candidate fragments required for progress records and statistical verification.
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, re-rank, filter, and upper-limit trimming). The progress table and statistical caliber description are then completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Project task book | Check whether the subject name, undertaking unit, leader, start and end dates, signature fields, and version number are consistent with the breakdown table | In the search test, switch searchMode, test once each with synonymous questions, record the number of candidates and scores, and compare the fields of the same record in the task book and the breakdown table |
| Subject task breakdown table | Check whether the task level, subject number, quantity, dates, and corresponding items in the task book are aligned item by item | Compare the order of the top few items before and after enabling usingReRank, and check whether the breakdown table candidates cover the subject entries in the task book |
| Budget book | Check whether the funding subjects, amount calibers, budget year, and the dates and quantities in the task book and breakdown table match | Adjust limit to observe changes in the number of candidates, separately record the length of the reference array, and check whether the tokens of the reference text retain key subject fragments |
| Signature pages and receipts | Check whether the signer, signing date, receipt number, and the versions of the task book, breakdown table, and budget book are synchronized | Search with the same question rephrased, check whether the signature page and receipt candidates appear together with the main record, and confirm that the version dates are consistent |
Three common errors
- Symptom: The name, quantity, or date of the same subject do not match between the task book and the breakdown table. Cause: Versions were not synchronized during the signing rounds, old drafts and new drafts were stored in the knowledge base at the same time, and retrieval candidates were mixed with fragments from different versions.
- Symptom: The budget book subjects and the task book funding calibers are inconsistent, and a discrepancy appears in the statistical total. Cause: The task book attachment was not updated after the budget adjustment, and the missing receipt made it impossible to verify the signing status.
- Symptom: There are many candidates, but the ones ranked first are all irrelevant, and the answer lacks key paragraphs. Cause: The retrieval mode does not match the material form, and after similarity filtering and reference upper-limit trimming, key fragments are truncated.
How to verify the setup
Acceptance of this function: test once each with synonymously rewritten questions, compare candidates and scores, then toggle re-ranking on and off to observe the ordering of the top few items; switch searchMode and adjust limit, record the number of candidates and scores, separately record the length of the reference array, and check the tokens of the reference text.
Business test: use one week of records to check the totals and see whether the calibers and exceptions are clearly written; compare the fields, dates, and signatures of the task book, breakdown table, budget book, and signing receipts item by item, and confirm that the number of records matches the statistical count.
- In the knowledge base search test, search the same question with different phrasings, record the number of candidates and similarity scores, and check whether different versions of the same record are recalled.
- Compare the order of the top few items before and after enabling
usingReRank, and check the tokens and array length of the reference text to confirm that key fragments have not been truncated or lost.