#contradiction benchmark

Sakana AI's LLM Peer Review System Catches 73% of Core-Claim ErrorsNEW⭐8

Sakana AI's Multi-Layered Review system catches 73% of core-claim errors in LLM peer review, outperforming baselines but failing on prompt injection.