为什么LLM在因果发现中失败以及干预代理如何逃脱
Why LLMs Fail at Causal Discovery and How Interventional Agents Escape
AI总结：
本文证明大型语言模型在因果发现中存在根本性失败，并提出一种基于干预代理的因果贝叶斯优化方法（A-CBO），通过外部贝叶斯循环在无需模型微调的情况下实现可证明的收敛。
摘要：
因果发现是科学推理的基石，但大型语言模型能否可靠地执行因果发现仍是一个悬而未决的问题。最近的基准测试表明，即使是微调后的模型在简单因果图上也会达到平台期，并随着复杂度增加而退化，但失败的原因尚未明确。我们证明这种失败是根本性的：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，任何这样做的尝试都需要模型的内部表示无限增长，从而违反了这些方法工作的条件。我们将其形式化为核障碍定理，确立该限制是学习范式固有的，而非任何特定模型或数据集。我们提出了代理因果贝叶斯优化（A-CBO），其中冻结的语言模型作为干预预言机，回答关于干预效果的目标查询，而外部贝叶斯循环在对数轮次内将信念集中在候选因果图上。由于决策在障碍适用的空间之外运行，A-CBO在底层模型保持不变的情况下可证明收敛。在Corr2Cause上，A-CBO无需任何训练即可匹配微调基线。在Extended Corr2Cause（一个扩展到24个变量、包含18K测试样本的新基准）上，A-CBO显著优于微调和偏好优化，且优势不断扩大。
英文摘要：
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, \emph{not any particular model or dataset}. We propose Agentic Causal Bayesian Optimization (A-CBO), wherein a frozen language model serves as an interventional oracle answering targeted queries about intervention effects, while an external Bayesian loop concentrates beliefs over candidate graphs in logarithmically many rounds. Because the decision operates outside the space where the obstruction applies, A-CBO provably converges while the underlying model remains unchanged. On Corr2Cause, A-CBO matches fine-tuned baselines without any training. On Extended Corr2Cause, a new benchmark scaling to 24 variables with 18K test samples, A-CBO significantly outperforms both fine-tuning and preference optimization, with the advantage growing
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
