Xiao Huang: Why was GPT-5.2 able to improve so significantly, even surpassing human experts?
DOORM: The core improvement of GPT-5.2 lies in optimizing its reasoning process through reinforcement learning, enabling the model to refine its thinking, try different strategies, and identify errors. This enhances logical reasoning capabilities, allowing the AI to analyze problems like humans rather than relying on rote memorization, thus excelling in professional tasks
.
发表回复
要发表评论，您必须先登录。