Knowledge Base Retrieval for School Exam Question Setting and Grading: Coordinate Operations and Application Handover

School Exams: review retrieval for operational handover. Compare the source records, required inputs and downstream results before handing the task over.

What this category's data looks like

This category's data includes exam question setting and marking materials such as the two-way specification table for question setting, exam papers, and marking rubrics, sourced from teaching and research groups and grade-level lesson preparation groups, with an update cadence that rolls by semester and exam batch. When entering the knowledge base retrieval and recall function, the specification table carries the mapping of assessment points and question types, the exam papers carry the original text of question stems and options, and the marking rubrics carry scoring points and deduction explanations. Field characteristics center on names, quantities, dates, and signatures: the name, score value, and question number of the same question in the specification table and the exam paper must be consistent, and the dates and signature receipts of the marking rubrics must be synchronized with the exam paper version. During retrieval, these materials enter as candidate fragments, for subsequent nodes or manual generation of handover-ready deployment status and operation records.

Configuration and implementation checks

This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate fragments and similarity scores (after mode, merge, rerank, filter, and limit truncation). The handover-ready deployment status and operation records are completed by subsequent nodes or manually.

ObjectCheckVerification action
Two-way specification table for question settingAssessment point names, question numbers, quantities, and dates must correspond item by item with the exam paper, and signatures must be completeIn the search test, test once each with synonymous questions, record the number of candidates and scores, and check whether the question numbers, score values, and dates of the specification table and the exam paper are consistent
Exam paperThe original text of question stems must be consistent with the scoring point names, quantities, and dates of the marking rubricsSwitch searchMode to semantic, full-text, or hybrid, compare whether candidates contain both semantic and full-text results, then toggle rerank on and off to see the order of the top few results
Marking rubricsScoring point names, quantities, dates, and signature receipts must be synchronized with the exam paper versionCheck the similarity scores of returned candidates and the number of items after limit truncation, and check whether the dates and signatures of the rubrics and the exam paper match
Deployment status and operation recordsVersion, connection, and task records must be consistent, and the handover form must be verifiable item by itemCheck item by item against the handover form once, see which record is missing, and record the changes in candidate ordering before and after toggling usingReRank

Three common errors

  • Symptom: The name, quantity, and date of the same question in the specification table and the exam paper do not match. Cause: The two materials were imported separately, versions were not synchronized, and retrieval candidates are mixed together.
  • Symptom: The marking rubrics lack signatures or receipts, and no corresponding record can be found on the handover form. Cause: Field verification was not performed before the materials entered retrieval, and the signature page was not updated with the version.
  • Symptom: The same question cannot be recalled when rephrased, and although there are many candidates, the ones ranked at the front are all irrelevant. Cause: The retrieval mode does not match the material form, rerank is not enabled, or similarity filtering is too strict.

How to verify the setup

Acceptance of this function: test once each with synonymously rewritten questions, compare candidates and scores, then toggle rerank on and off to see the ordering of the top few results; record the number of candidates under searchMode, similarity, usingReRank, and limit.

Business testing: check item by item against the handover form once, see which record is missing; check whether the names, quantities, dates, and signatures of the specification table, exam paper, and marking rubrics are consistent.

  • In the knowledge base search test, switch searchMode and compare the number of candidates and similarity scores of the semantic, full-text, and hybrid modes.
  • Toggle rerank on and off, compare the order of the top few candidates, and record the cited text tokens and array item count after usingReRank and limit truncation.