返回全部动态
别被迷惑：大语言模型并不会推理
原标题：Don’t be fooled—LLMs don’t reason
AI 摘要
前Google DeepMind研究员撰文指出，当前大语言模型并不具备真正的推理能力，其链式思维仍由同一套下一词预测机制生成，缺乏可检查的持久认知状态、知识与推理的分离，以及真实的推理过程。作者以AlphaGo的搜索机制为例，认为真正的推理应像AlphaGo那样维护显式的博弈树并据此更新判断，并透露自己已离开Google DeepMind，希望推动借鉴AlphaGo架构的新推理方法。
以上摘要由 AI 生成，可能存在误差。事实请以原文为准。
正文节选
On an afternoon in Seoul in March 2016, I watched a program I helped build put a stone on the fifth line of a Go board in what looked like a gift to its human opponent. Move 37 in game two of the five-game match looked so absurd that some commentators thought it was a programming glitch. It wasn’t. AlphaGo won the game, ultimately triumphing 4-1 over Lee Sedol, one of the greatest professional Go players of all time. “I thought AlphaGo was based on probability calculation and that it was merely
发布时间：2026-10-02 16:00
抓取时间：2026-10-02 16:56
来源机构：MIT Technology Review