Tag
The paper compares five LLM-based review paradigms across three input representations and three LLMs on 232 ICLR 2025 submissions, finding that direct rating matches comparative ranking after recalibration with far fewer API calls, abstract-only input beats full-paper summaries, and human scores still predict citations better than any LLM paradigm except Group Rating.
This paper studies the feedback loop where AI-generated reviews influence future training of AI reviewers, leading to reduced judgment diversity called 'scientific-judgment collapse,' and introduces TrustReviewer, an open-source system to mitigate this through curated training and activation steering.