PremiseCheck Framework for Efficient Peer Review Fact-Checking

PremiseCheck improves fact-checking efficiency by using an LLM to decompose review comments into explicit premises, which are then verified by low-cost decision models. This two-stage approach achieves accuracy comparable to frontier models at half the cost and significantly lower latency.