Sakana AI’s AI model Peer Review System Catches 73% of Core-Claim Errors

What happened

Sakana AI’s TMLR paper introduces Multi-Layered Review, a 3-agent Claude-based reviewer, and a 1,164-error Contradiction Benchmark. Sakana AI has published Beyond Imitation , a TMLR research paper on LLM (the kind of AI system trained on text to produce text)-assisted peer review built around error detection.

The research team ships two pieces: a Contradiction Benchmark and a Multi-Layered Review (MLR) system. Most AI reviewers are graded on how closely they copy human reviews.

For developers building research agents, the lesson is practical. TL;DR Size: 1,164 inserted contradictions across 257 papers from 5 venues.