What this category's data looks like
This page focuses on graduate training process management, and the materials come from three types of process documents: training plans, proposal reports, and mid-term assessment forms. The training plan is jointly confirmed by the supervisor and the graduate student, and includes course credits, research direction, and time nodes; the proposal report records the basis for topic selection, literature review, and argumentation arrangements; the mid-term assessment form summarizes stage results, existing problems, and follow-up plans. The three types of materials play different roles in Q&A acceptance: the training plan provides the training framework and progress baseline, the proposal report supports topic selection and argumentation-type questions, and the mid-term assessment form reflects stage execution and deviations. The update rhythm advances with the semester, with submissions in batches at the proposal and mid-term nodes, and field characteristics are concentrated in four categories: name, quantity, date, and signature. The same record is prone to differences across materials in name wording, quantity definitions, date order, and signature receipts, and these must be compared item by item during acceptance.
Configuration and implementation checks
This part is executed by the memory turns and prompt configuration in the application settings, and outputs the historical message range and answer constraints to reference. The acceptance records and the completed follow-up questions are handled by the model, while the external entry point and authorization are handled by the publishing service and file permissions. Memory turns control how many previous dialogue turns can be referenced at most when answering; if the value is too small, the preceding context cannot be used, and if the value is too large, irrelevant content is brought in and consumption increases.
| Object | Check | Verification action |
|---|---|---|
| Training plan | Whether the name, quantity, date, and signature correspond item by item with the proposal report and mid-term assessment form | Check the memory turns value in the application settings, run a trial with two consecutive questions, check whether the second round can carry forward the reference from the first question, and compare the preceding context range in messages |
| Proposal report | Whether the name, quantity, and date of the same record are consistent across the two materials, and whether the versions are synchronized | Run a trial with two consecutive questions, observe whether the second round references the material fields already given in the first round, and check whether the historical messages under chatId are correctly included |
| Mid-term assessment form | Whether missing signatures or receipts are identified and followed up on, and whether the date order conflicts with the training plan | Ask a question that lacks key information, see whether it follows up first before answering, and record the completed follow-up questions and source materials in variables |
| Acceptance records and completed follow-up questions | Missing items are identified and followed up on, and answers are based on materials | Compare the usage records of two runs, confirm that consumption changes accordingly after adjusting the turns, and check whether the acceptance record is forwarded to the handler |
Three common errors
- Phenomenon: the second question answers off-topic when using a reference. Cause: the memory turns value does not match the preceding context range required by the question, so the preceding context is not included for reference.
- Phenomenon: the name, quantity, and date of the same record do not match between the training plan and the proposal report. Cause: the versions are not synchronized, and the fields were not compared item by item during acceptance.
- Phenomenon: a mid-term assessment form missing a signature or receipt is still accepted as complete material. Cause: missing-item identification did not trigger a follow-up, and the completed follow-up questions were not recorded.
How to verify the setup
This feature acceptance: ask two consecutive follow-up questions, see whether the second question carries forward the reference from the first question, and check whether the preceding context range in messages matches the memory turns value.
Business test: ask a question that lacks key information, see whether it follows up first before answering, and whether the follow-up items fall into variables and are forwarded to the handler.
- Check the current memory turns value in the application settings, run a trial with two consecutive questions, and confirm that the second round can correctly carry forward the reference from the first round.
- Compare the usage records of two runs, confirm that consumption changes accordingly after adjusting the turns, and check that the opening message is not treated as preceding context.