Meta’s SPICE framework lets AI systems teach themselves to reason
Researchers at Meta FAIR and the National University of Singapore have developed a new reinforcement learning framework for self-improving AI systems. Called Self-Play In Corpus Environments (SPICE), ...
