Rohan Paul (@rohanpaul_ai) · 2026/10/22 13:27:55
Meta 与 MSU 研究：LLM 构建推理图质量波动大且误差传导
Meta 与密歇根州立大学最新论文指出，由大语言模型自动构建的推理图（Reasoning Graphs）存在显著的质量差异，且这些结构性错误会直接污染基于其计算的置信度分数。研究发现，推理图的质量更多取决于模型家族和图结构形态，而非单纯的参数规模，甚至出现小模型优于大模型的异常现象。该结论警示开发者在使用 LLM 生成子问题分解图时，必须对图本身进行独立验证，不能盲目信任下游答案或置信度。
报道全文原始报道全文
New Meta + Michigan State Univ paper finds that LLM-built reasoning graphs vary widely in quality, and their errors carry into the confidence scores computed from them. Many methods split a question into a graph of sub-questions, then compute answers or confidence on top. They assume the LLM's graph is correct, which is rarely checked. They tested 9 models with less and less help, from wiring given steps to building the whole graph. Quality tracked model family and graph shape more than size, and Qwen3-32B answered sub-questions worse than Qwen3-8B. If you build on LLM-made reasoning graphs, check the graph itself before trusting answers or confidence computed from it.