arXiv AI recent: LLM-as-Judge in Education: A Curriculum-Grounded Marking Pipeline
Researchers co-developed a curriculum-grounded LLM-as-Judge pipeline for question-level marking to support university admission exam preparation. The system uses a staged workflow to gene...
The pipeline identifies topics, subtopics, and cognitive demand to assemble verifiable context for LLM judgment. It operationalizes curriculum intent using syllabus artifacts such as prescribed verbs, outcomes, performance band descriptors, glossary definitions, and marking-guideline principles.