ConceptualGrader: Interactive Provenance and Eligibility Techniques for Auditing Large Language Models in Grading

Under review. An empirical study showing that LLMs with similar grading accuracy impose sharply different human review burdens, paired with a system that makes the model's grading eligibility, cited evidence, and override records inspectable.

Authors: Iman YeckehZaare · Venue/status: UIST

This submitted manuscript shows that models with comparable short-answer grading accuracy produce false-positive review queues that differ several-fold — a governance gap that accuracy alone hides — and introduces ConceptualGrader, in which instructors set which answers the model may judge, review the model's cited evidence before accepting credit, and preserve corrections for reuse.

Public artifact boundary: this route exposes status, authorship, visual summary, citation metadata, and the contribution boundary; manuscript files are posted only when review and prepublication rules allow it.

The rendered manuscript page adds status, visual summary, review boundary, citation metadata, and contribution notes. Key links: home, systems, papers, manuscripts, Google Scholar, GitHub, LinkedIn, ORCID, MIT profile, and CV PDF.