What this category's data looks like
Research project annual progress and acceptance materials cover three categories: annual progress reports, mid-term review materials, and acceptance reports. Annual progress reports are submitted by cycle and record phase tasks, indicators, and personnel; mid-term review materials focus on milestone comparison and adjustment notes; acceptance reports summarize full-cycle results, funding, and signed receipts. Sources are the project team's multilingual original records and translated drafts, updated on a rolling basis with review milestones. Field characteristics: the name, quantity, date, and signature of the same record must be consistent across multiple materials, and translated names must refer back to the original text. When materials enter document parsing and chunking, annual progress reports and mid-term review materials participate in initial training as data rows to be indexed, and acceptance reports are used for retrieval and subsequent processing after indexing is complete to generate a unified terminology table and comparison draft.
Configuration and implementation checks
Parsing and training tasks (server-side tasks) only output data rows and index status (parsed, indexing, indexed); the unified terminology table and comparison draft must be handled next by retrieval and subsequent processing. After parsing is complete, parsed chunks are immediately persisted as visible data rows and bound to a unique data identifier, and indexing flows through three states; data to be indexed does not participate in recall, and changing models only rebuilds data that has completed indexing.
| Object | Check | Verification action |
|---|---|---|
| Annual progress report | The name, quantity, and date of the same record must be consistent in the mid-term review materials and acceptance report, and signatures and receipts must not be missing | Check whether the data row status is parsed / indexing / indexed, and check whether the chunks corresponding to chunkSize and chunkSplitter cover the record |
| Mid-term review materials | The version must be synchronized with the annual progress report, and the translated name of the same term must be consistent throughout and able to refer back to the original text | Check the file record and data row identifier separately, confirm the training mode corresponding to trainingType, and sample three terms to check whether their wording is consistent throughout |
| Acceptance report | Fields, dates, and signatures are complete, and the name, quantity, and date of the same record match across the two materials | Check whether the data rows covered by the rebuild task contain only data that has completed indexing, and check the signature and receipt fields |
| Unified terminology table and comparison draft | Terminology is consistent throughout, translated names can refer back to the original text, and the number of citations and the tokens of cited text are recorded separately | Sample three terms to check whether their wording is consistent throughout, check that the number of citations is the array length, and check the citation content budget against the tokens corresponding to quoteMaxToken |
Three common errors
- Symptom: the name, quantity, and date of the same record do not match between the annual progress report and the acceptance report. Cause: multilingual materials have not unified terminology, and after chunking each material is persisted independently, lacking cross-material comparison.
- Symptom: the version of the mid-term review materials is not synchronized, and signatures or receipts are missing. Cause: after materials are updated they are not re-parsed and persisted, old data rows are still in the completed indexing state, and the retrieval layer recalls the old version.
- Symptom: newly uploaded materials cannot be found, and failed training leaves empty data rows. Cause: data rows in the to-be-indexed state do not participate in recall, interrupted training leaves orphan data, and retry or deletion followed by re-parsing is required.
How to verify the setup
Acceptance of this feature: upload one material, observe the data row status transitions and the coverage scope of the rebuild task, and confirm that after parsing is complete it is immediately persisted as visible data rows, indexing flows through the three states parsed / indexing / indexed, and changing models only rebuilds data that has completed indexing.
Business test: sample three terms to check whether their wording is consistent throughout, check whether translated names can refer back to the original text, and confirm that the number of citations is the array length and the citation content budget is recorded separately according to the tokens corresponding to quoteMaxToken.
- Open the data row to check the status, confirm it is consistent with the interface display, and check the file record and data row identifier separately.
- Check the data rows covered by the rebuild task, and confirm that they contain only data that has completed indexing and that data to be indexed does not participate in the rebuild chain.