Sakana AI's self-improving model wins nearly 70% of matchups
Sakana AI's MASS method hit a nearly 70% win rate against its base model after two rounds — but errors shared across all three of its roles get reinforced, not caught.
Topic
Topic
Sakana AI's MASS method hit a nearly 70% win rate against its base model after two rounds — but errors shared across all three of its roles get reinforced, not caught.