What this category's data looks like
In business basis queries, university admissions physical examination organizations use three types of materials: physical examination forms, re-examination applications, and restricted-major notifications. Physical examination forms come from physical examination institutions and are updated by batch; their fields include name, examination items, conclusion, date, and signature. Re-examination applications come from candidates or parents and are updated by application node; their fields include application reason, original examination conclusion, signature, and receipt. Restricted-major notifications come from admissions departments and are updated by annual admissions policy; their fields include major name, restriction basis, notification date, and acknowledgment. After these three types of materials enter parsing and training tasks, each becomes an independent data row. Physical examination forms are used to check the conclusion and date, re-examination applications are used to check the application and receipt, and restricted-major notifications are used to check the major name and restriction entries. In terms of update cadence, physical examination forms are stored by batch, re-examination applications are stored as applications are submitted, and restricted-major notifications are stored with the annual policy; all must be bound to a unique data identifier.
Configuration and implementation checks
This step is executed by parsing and training tasks (server-side tasks), outputting data rows and index status (parsed, indexing, indexed). Basis entries that can be traced back to the original text are then completed by retrieval and subsequent processing. After parsing is complete, parsing chunks are immediately stored as visible data rows and bound to a unique data identifier. The index transitions through three states, and data pending indexing does not participate in recall.
| Object | Check | Verification action |
|---|---|---|
| Physical examination form | Fields, date, and signature must match the original; the conclusion must correspond to the major name and restriction basis in the restricted-major notification | Open the data row to check the status and original text, and verify the name, examination items, conclusion, date, and signature; use chunkSize and chunkSplitter to check whether the chunks cover the conclusion section |
| Re-examination application | Application reason, original examination conclusion, signature, and receipt must be complete; the original examination conclusion must match the name and date of the corresponding record in the physical examination form | Open the data row to check the status and original text, and verify the application reason, original examination conclusion, signature, and receipt; record the number of cited entries and cited text tokens; quoteMaxToken is only the cited content budget |
| Restricted-major notification | Major name, restriction basis, notification date, and acknowledgment must match the annual policy; the restriction basis must be traceable back to the physical examination form conclusion | Open the data row to check the status and original text, and verify the major name, restriction basis, notification date, and acknowledgment; use trainingType to distinguish Vector and QA processing modes, and view the training progress by category |
| Data row status | It is in one of the three states parsed, indexing, or indexed and matches the interface display; data pending indexing does not participate in recall | Open the data row to check the status; after changing models, check that the rebuild task covers only indexed data rows, and data pending indexing does not participate in the rebuild chain |
Three common errors
- Symptom: The name and date of the same record do not match between the physical examination form and the re-examination application. Cause: The two materials are each stored as independent data rows, lacking cross-material traceback verification.
- Symptom: The major name in the restricted-major notification is not synchronized with the version of the physical examination form conclusion. Cause: After the annual policy update, the old data row is still indexed and has not been rebuilt with the new version.
- Symptom: The re-examination application lacks a signature or receipt, but retrieval can still hit it. Cause: Successful parsing does not mean retrieval will definitely hit; the index status and field completeness must also be checked.
How to verify the setup
Acceptance for this feature: upload one material and observe the data row status transitions and the coverage scope of the rebuild task. The status should be one of the three states parsed, indexing, or indexed and should match the interface display; after changing models, the rebuild task covers only data rows whose indexing is complete, and data pending indexing does not participate in the rebuild chain.
Business test: ask a cross-material basis question and see whether the answer provides a traceable source. For example, ask which physical examination form conclusion the restriction basis in the restricted-major notification corresponds to, and see whether the answer provides basis entries that can be traced back to the original text, and record the number of cited entries and cited text tokens; quoteMaxToken is only the cited content budget.
- Open the data row to check the status and original text, verify the identifiers of the file record and the data row separately, and confirm that the identities of the file record and the data row are distinguishable.
- View the data rows covered by the rebuild task, confirm that only data whose indexing is complete is covered; check the chunking and processing mode parameters corresponding to the three identifiers
chunkSize,chunkSplitter, andtrainingType.