Sakana AI is an AI research company based in Tokyo. Recent coverage has centered on self-improvement work: its MASS method, in which a single language model improves itself and reaches a nearly 70% win rate against the base model after two rounds; an LLM-based peer review system that catches 73% of core-claim errors; and the hiring of Jürgen Schmidhuber as Chief Scientific Advisor to lead its new RSI Lab.
Sakana AI's self-improving model reaches 70% win rate but has blind spotsSakana AI's MASS method lets a single language model improve itself by solving tasks, judging results, and refining workflows, reaching a nearly 70% win rate against the base model after two rounds. However, if the model makes the same errors across all three roles, it can mistake them for progress and reinforce them during training.
Sakana AI Peer Review System Catches 73% of Core-Claim ErrorsSakana AI has introduced an LLM-based peer review system that catches 73% of core-claim errors in papers, potentially assisting academic review and improving efficiency and quality.
Sakana AI Hires Deep Learning Pioneer SchmidhuberTokyo-based Sakana AI hired Jürgen Schmidhuber as Chief Scientific Advisor. He will help lead its new RSI Lab, focused on recursive self-improvement. His 1990s ideas have already influenced projects like the Darwin Gödel Machine, and Sakana calls him the father of modern AI.
The A.I. Daily