[论文] Why LLMs Fail at Causal Discovery and How Interventional Agents E...

论文概要

研究领域: AI
作者: Amartya Roy, Sonali Parbhoo
发布时间: 2026-05-28
arXiv: 2605.27567

中文摘要

因果发现是科学推理的基石,但大语言模型能否可靠执行因果发现仍悬而未决。近期基准测试显示,即便是微调模型也在简单因果图上遇到瓶颈,且随复杂度增加而退化--但其失败根源尚未厘清。本文证明这一失败是根本性的:监督微调、直接偏好优化和上下文学习都会产生无法区分生成相似观测数据的因果图的预测器,且任何试图区分它们的尝试都要求模型内部表示无界增长,从而违反这些方法赖以成立的前提条件。作者将其形式化为核障碍定理,证明该限制内在于学习范式本身,而非特定模型或数据集。为此,提出了智能体因果贝叶斯优化(A-CBO):冻结的语言模型作为干预预言机回答针对性查询,外部贝叶斯循环在log轮次内集中对候选图的信念。由于决策在障碍适用的空间之外运作,A-CBO可被证明收敛,而底层模型保持不变。在24变量、18K测试样本的Extended Corr2Cause新基准上,A-CBO显著优于微调和偏好优化基线。

原文摘要

Causal discovery is a cornerstone of scientific reasoning, yet whether large language models can perform it reliably remains an open question. Recent benchmarks show that even fine-tuned models plateau on simple causal graphs and degrade as complexity grows, but why they fail has not been established. We prove the failure is fundamental: supervised fine-tuning, direct preference optimization, and in-context learning all produce predictors that cannot distinguish between causal graphs generating similar observational data, and any attempt to do so requires the model's internal representations to grow unboundedly, violating the very conditions under which these methods work. We formalize this as a kernel obstruction theorem, establishing that the limitation is intrinsic to the learning paradigm, not any particular model or dataset. We propose Agentic Causal Bayesian Optimization (A-CBO), w...

自动采集于 2026-05-29

#论文 #arXiv #AI #小凯