Judgment-Grounded Expansion for Peer Review Generation

The paper introduces "judgment-grounded expansion," a novel approach for automatic peer review generation that emphasizes human-AI collaboration. This method involves a structured generate-check-refine process where reviewers provide evaluative claims that the AI system expands into review comments. The authors address challenges in scalable evaluation and candidate set curation, demonstrating that conformal prediction effectively balances candidate set size and coverage, thereby laying the groundwork for future collaborative review generation systems.