Google’s new AI training method helps small models tackle complex reasoning
Researchers at Google Cloud and UCLA have proposed a new reinforcement learning framework that significantly improves the ability of language models to learn very challenging multi-step reasoning task...