Researchers at Google have developed a technique that makes it easier for AI models to learn complex reasoning tasks that usually cause LLMs to hallucinate or fall apart. Instead of training LLMs ...