What this category's data looks like
This page addresses material verification and discrepancy localization in graduate training process management. The materials entering this function include training plans, proposal reports, and mid-term assessment forms, which come from supervisors, school academic affairs, and assessment panels respectively, and are updated according to semester milestones. A training plan contains course names, credit counts, start and end dates, and supervisor signatures; a proposal report contains the thesis title, proposal date, review panel signatures, and receipt; a mid-term assessment form contains the assessment conclusion, assessment date, and assessment panel signatures. When these three types of materials enter the tool calling node, each serves as an independent data source, and the model determines whether to call tools to read fields, compares the names, quantities, and dates of the same object across two materials item by item, locates inconsistent fields, and forms the original call records of the discrepancy list for subsequent reviewers to confirm and write back.
Configuration and implementation checks
This step is executed by the tool calling node in the workflow (the model decides whether to call, and the node executes), outputting the tool call results (the inputs and returns in the call records). The discrepancy list is then completed by subsequent AI nodes or manual verification.
| Object | Check | Verification action |
|---|---|---|
| Training plan | Whether the course names, credit counts, start and end dates, and supervisor signatures are consistent with the corresponding fields in the proposal report and mid-term assessment form | Open the tool node introduction fields and check whether their purpose and returns can be understood; run a test with a question missing a required parameter and observe whether it asks a follow-up question first |
| Proposal report | Whether the thesis title, proposal date, review panel signatures, and receipt are consistent with the corresponding fields in the training plan and mid-term assessment form | Open each parameter configuration and check the parameter introduction and whether it is required; check the diamond connectors and node status on the canvas |
| Mid-term assessment form | Whether the assessment conclusion, assessment date, and assessment panel signatures are consistent with the corresponding fields in the training plan and proposal report | Check the inputs and returns in the tool call records, and confirm that the returned results do not contain values that do not exist in the business materials |
| Tool calling node | The model decides whether to call based on the tool introduction, parameter introduction, and whether it is required; when there are no required parameters, it may fabricate parameter values | Run a test with a question missing that field and see whether it asks a follow-up question first; check the values of toolChoice and functionCall in the call records |
Three common errors
- Symptom: The credit counts for the same record do not match between the training plan and the mid-term assessment form. Cause: The versions of the two materials are not synchronized, and the tool call only reads the field value from one of them.
- Symptom: The proposal dates in the proposal report and the mid-term assessment form are inconsistent. Cause: The two materials are filled in by different roles, the date fields are not compared across materials, and the discrepancy is not localized.
- Symptom: The mid-term assessment form is missing the assessment panel signatures or receipt. Cause: The tool call return result does not contain that field, the model does not ask a follow-up question, and directly gives an answer.
How to verify the setup
Acceptance for this function: run a test with a question missing a required parameter, check the inputs and returns in the call records, confirm whether it asks a follow-up question or calls according to the introduction, and check toolChoice, functionCall, and the connection status of the tool calling node.
Business test: use a real record to find a field that is inconsistent with the ledger, and confirm that the discrepancy list can localize the specific location where the name, quantity, or date does not match.
- Open the introduction field in the tool node, check whether this description can be understood when taken out alone, and check the parameter introduction and whether it is required.
- On the canvas, check that the diamond interface is connected to the target tool, check the node status, and run a test with a question missing that field to see whether it asks a follow-up question first.