Comparative Evaluation of LLM-Based Review Paradigms for Academic Paper Quality Assessment

Growing submission volumes have renewed interest in using large language models (LLMs) to support peer review. It remains unclear how such evaluation should be structured, how much the model matters, and whether LLM scores predict research impact. We evaluate five paradigms across three input repres