CauTion: Knowing When to Trust LLMs for Ensemble Causal Discovery
2026-06-02 • Machine Learning
Machine LearningArtificial IntelligenceComputation and Language
AI summaryⓘ
The authors introduce CauTion, a method that combines knowledge from large language models (LLMs) with several statistical causal discovery algorithms to better find cause-and-effect relationships from data. Their approach first agrees on clear connections using multiple algorithms, then smartly decides when to trust the LLM to resolve uncertain cases, and finally ensures the resulting cause-effect map is logically valid. Tests show CauTion works better than existing methods, especially on bigger problems, and is careful about LLM mistakes. This is done without needing extra labeled data to judge trust.
causal discoveryobservational datalarge language modelsensemble methodsconsensus votingtrust calibrationcausal graphacyclicitystatistical inferencealgorithmic bias
Authors
Bo Peng, Kaiwen Wu, Sirui Chen, Zhiheng Wang, Yu Qiao, Chaochao Lu
Abstract
Causal discovery from observational data remains challenging due to the fundamental limitations of purely statistical methods, such as statistical distinguishability within equivalence classes and sensitivity to finite sample sizes. While large language models (LLMs) offer a promising source of domain knowledge to complement statistical inference, existing LLM-augmented methods are vulnerable to LLM errors and incur high token costs. Moreover, reliance on a single data-centric algorithm can make results sensitive to algorithm-specific biases. To address these limitations, we propose CauTion, a framework that reliably integrates LLM domain knowledge into an ensemble of statistical causal discovery algorithms through consensus filtering and LLM reliability estimation. CauTion proceeds in three stages. First, an algorithm ensemble utilizes a consensus voting to resolve up to 96% of edges on which algorithms agree, achieving near-perfect accuracy on the filtered consensus edges. Second, a trust-calibrated arbitration mechanism estimates the relative reliability of the LLM and the algorithms via an annotation-free trust calibration procedure, which is then utilized to govern a trust-weighted voting process that restricts LLM arbitration exclusively to edges with unreliable algorithmic evidence. Third, a cycle repair step is applied to guarantee the final causal graph is validly acyclic. Experiments on six datasets demonstrate that CauTion consistently outperforms both data-centric and LLM-augmented baselines, with larger gains on larger graphs and strong robustness to LLM errors. Code is available at https://github.com/OpenCausaLab/CauTion.