Why LLMs Fail at Causal Discovery and How Interventional Agents Escape

• URL: http://arxiv.org/abs/2605.27567v1
• Date: Tue, 26 May 2026 18:37:03 GMT
• Title: Why LLMs Fail at Causal Discovery and How Interventional Agents Escape
• Abstract summary: Causal discovery is a cornerstone of scientific reasoning, yet whether large language models can perform it reliably remains an open question.<n>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.<n>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.
• Score: 2.7976995580303305
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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

Related papers

• Vision: Data-Centric Anchoring for Robust and Interpretable Agentic AI [21.406298170489507]
  We argue that robustness and interpretability should be engineered into the data environment, not extracted from models after training.<n>Our central contribution is the Data-Centric Agentic Loop, a four-stage framework of Curate, Augment, Constrain, and Attribute.<n>We ground the framework in a failure-driven taxonomy that links four core failure modes to the data lifecycle.
  arXiv  Detail & Related papers  (2026-09-08T03:57:54Z)
• Inference-Time Conformal Reasoning with Valid Factuality Control for Large Language Models [19.716905963081956]
  We propose an emphInference-Time Conformal Reasoning framework.<n>It learns a structure-level factuality uncertainty function that aggregates claim-level factuality signals over reasoning graphs.<n>We theoretically show such generation is nested, yielding valid coverage guarantees for factuality control.
  arXiv  Detail & Related papers  (2026-06-07T20:52:10Z)
• Why Do Reasoning Models Lose Coverage? The Role of Data and Forks in the Road [26.088108937790167]
  We investigate the reasoning shrinkage arise under SFT-based post-training.<n>By tracking post-training dynamics in settings, we find that the shrinkage phenomenon is tightly correlated with the prevalence of decision-point scenarios in the training data.<n>Our findings identify data-centric factors as a key driver of shrinkage in reasoning models and highlight diversity-aware designs as an effective lever for controlling it.
  arXiv  Detail & Related papers  (2026-05-16T14:55:12Z)
• GTS: Inference-Time Scaling of Latent Reasoning with a Learnable Gaussian Thought Sampler [54.10960908347221]
  We model latent thought exploration as conditional sampling from learnable densities and instantiate this idea as a Gaussian Thought Sampler (GTS)<n>GTS predicts context-dependent perturbation distributions over continuous reasoning states and is trained with GRPO-style policy optimization while keeping the backbone frozen.
  arXiv  Detail & Related papers  (2026-02-15T09:57:47Z)
• Causal Bayesian Optimization with Unknown Graphs [17.871251114630827]
  Causal Bayesian Optimization (CBO) is a methodology designed to optimize an outcome variable by leveraging known causal relationships through targeted interventions.<n>Traditional CBO methods require a fully and accurately specified causal graph, which is a limitation in many real-world scenarios where such graphs are unknown.<n>We propose a new method for the CBO framework that operates without prior knowledge of the causal graph.
  arXiv  Detail & Related papers  (2025-03-25T11:14:37Z)
• DeCaFlow: A deconfounding causal generative model [48.86777554649186]
  We introduce DeCaFlow, a deconfounding causal generative model.<n>We extend previous results on causal estimation under hidden confounding to show that a single instance of DeCaFlow provides correct estimates for all causal queries identifiable with do-calculus.<n>Our empirical results on diverse settings show that DeCaFlow outperforms existing approaches, while demonstrating its out-of-the-box applicability to any given causal graph.
  arXiv  Detail & Related papers  (2025-03-19T11:14:16Z)
• Causal Invariance Learning via Efficient Optimization of a Nonconvex Objective [12.423111378195667]
  We propose a novel method for finding causal relationships among variables.<n>We show that our method converges to a standard gradient method.<n>Our algorithm avoids exhaustive searches, making it especially when the number of covariants is large.
  arXiv  Detail & Related papers  (2024-12-16T15:11:02Z)
• Model-based Causal Bayesian Optimization [78.120734120667]
  We propose model-based causal Bayesian optimization (MCBO) MCBO learns a full system model instead of only modeling intervention-reward pairs. Unlike in standard Bayesian optimization, our acquisition function cannot be evaluated in closed form.
  arXiv  Detail & Related papers  (2022-11-18T14:28:21Z)
• Active Bayesian Causal Inference [72.70593653185078]
  We propose Active Bayesian Causal Inference (ABCI), a fully-Bayesian active learning framework for integrated causal discovery and reasoning. ABCI jointly infers a posterior over causal models and queries of interest. We show that our approach is more data-efficient than several baselines that only focus on learning the full causal graph.
  arXiv  Detail & Related papers  (2022-06-04T22:38:57Z)
• Variational Causal Networks: Approximate Bayesian Inference over Causal Structures [132.74509389517203]
  We introduce a parametric variational family modelled by an autoregressive distribution over the space of discrete DAGs. In experiments, we demonstrate that the proposed variational posterior is able to provide a good approximation of the true posterior.
  arXiv  Detail & Related papers  (2021-06-14T17:52:49Z)
• Causal Expectation-Maximisation [70.45873402967297]
  We show that causal inference is NP-hard even in models characterised by polytree-shaped graphs. We introduce the causal EM algorithm to reconstruct the uncertainty about the latent variables from data about categorical manifest variables. We argue that there appears to be an unnoticed limitation to the trending idea that counterfactual bounds can often be computed without knowledge of the structural equations.
  arXiv  Detail & Related papers  (2020-11-04T10:25:13Z)
• An Analysis of the Adaptation Speed of Causal Models [80.77896315374747]
  Recently, Bengio et al. conjectured that among all candidate models, $G$ is the fastest to adapt from one dataset to another. We investigate the adaptation speed of cause-effect SCMs using convergence rates from optimization. Surprisingly, we find situations where the anticausal model is advantaged, falsifying the initial hypothesis.
  arXiv  Detail & Related papers  (2020-05-18T23:48:56Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.