GraphReview: Scientific Paper Evaluation via LLM-Based Graph Message Passing
TL;DR AI
2 min readKey summary
Researchers introduced GraphReview, an LLM-based graph framework for scientific paper evaluation.
It represents papers and their relationships as a semantic graph, using LLMs to score nodes and compare papers across edges.
The system propagates review signals through related and prior work, improving paper ranking, decision prediction, and review generation.
Reported results show strong gains over baseline methods on ranking, acceptance decision, and review-quality tasks.