What this category's data looks like
This page addresses the scenario of exam question setting and grading in primary and secondary schools. Materials include the question-setting two-way specification table, exam papers, scoring rubrics, as well as signable material drafts and handover records. These materials come from teaching and research groups and question-setting groups, and are updated according to exam milestones. Their field characteristics center on name, quantity, date, version, and signature. When the question-setting two-way specification table enters this feature, it assumes the role of structural framework; the exam paper assumes the role of content subject; the scoring rubric assumes the role of scoring basis; and the drafts and handover records assume the role of circulation vouchers. After materials enter this feature, their respective fields and responsible persons must first be clarified, then they enter the model access and configuration stage, where the server side completes model binding and channel responses, and dialogue and retrieval complete the drafts and handover records accordingly.
Configuration and implementation checks
A. This step is executed by the server-side model channel configuration, outputting model binding relationships and channel responses. The signable material drafts and handover records are then completed by subsequent AI nodes, code nodes, or manual work. Models are divided into five categories: language, index, rerank, speech synthesis, and speech recognition. Different steps bind different model types.
| Object | Check | Verification action |
|---|---|---|
| Question-setting two-way specification table | Consistent with the name, quantity, date, version, and signature of the exam paper and scoring rubric | Check the name, question count, date, and version number of the three materials, and verify whether signatures and receipts are complete; against the model name returned by the channel, confirm that the language model and index model are bound correctly |
| Exam paper | Consistent with the fields and date of the question-setting two-way specification table and scoring rubric | Check item by item the question numbers, question types, scores, and dates of the exam paper and the specification table, and verify whether the versions are synchronized; confirm that the index model is used for retrieval and the rerank model is used to optimize retrieval ranking |
| Scoring rubric | Consistent with the name, quantity, and date of the exam paper and question-setting two-way specification table | Check the question numbers, scores, and dates of the scoring rubric and the exam paper, and verify signatures and receipts; confirm that the language model is used for dialogue and the index model is used for text block indexing |
| Model configuration | The model type corresponds to the step in which it is to be used, and the channel and key sources are consistent | In the model configuration, check the type labels and compare them against the model name returned by the channel; check the binding relationships of maxContext, maxTokens, and toolChoice, and confirm that quoteMaxToken is the quoted content budget, with the number of quotes recorded separately |
Three common errors
- Symptom: The name, quantity, and date of the same record do not match between the question-setting two-way specification table and the exam paper. Cause: During drafting, checks were not performed against unified fields, versions were not synchronized, and signatures or receipts were missing before handover.
- Symptom: The question numbers, scores, and dates of the scoring rubric and the exam paper are inconsistent. Cause: The two materials were drafted separately, did not go through the same handover record, and the lack of receipts allowed discrepancies to go undetected.
- Symptom: Retrieval reports index-model-related errors, or the answer style changes abruptly after switching models. Cause: The model type is inconsistently bound to the step, the channel and key sources do not match, and authentication fails.
How to verify the setup
Feature acceptance: Run one dialogue and one retrieval, check the model identifiers in the responses, and confirm that the language model and index model bindings match the names listed in the documentation; check whether maxContext, maxTokens, and toolChoice are configured according to the step.
Business test: Draft one document according to the template and complete one handover, and see whether missing items are blocked.
- In the model configuration, check that the model type labels correspond to the step in which they are to be used, and compare them against the model name returned by the channel.
- Check the channel configuration and key sources, confirm that authentication succeeds; check that
quoteMaxTokenis the quoted content budget, with the number of quotes recorded separately.