Why LLMs Fail at Causal Discovery and How Interventional Agents Escape
This paper proves that large language models have a fundamental limitation in performing causal discovery: methods like supervised fine-tuning, direct preference optimization, and in-context learning cannot distinguish between causal graphs that generate similar observational data. The authors propose Agentic Causal Bayesian Optimization (A-CBO), where a frozen language model serves as an interventional oracle and an external Bayesian loop converges to candidate graphs in logarithmically many rounds. On Corr2Cause, A-CBO matches fine-tuned baselines without any training; on Extended Corr2Cause (scaling to 24 variables and 18K test samples), A-CBO significantly outperforms both fine-tuning and preference optimization.
Article intelligence
Key points
- Proves that LLM failure in causal discovery is fundamental, due to a kernel obstruction theorem
- Proposes A-CBO, combining a frozen LLM with external Bayesian optimization
- Matches fine-tuned baselines on Corr2Cause without training; significantly outperforms on extended benchmark
Why it matters
This matters because proves that LLM failure in causal discovery is fundamental, due to a kernel obstruction theorem.
Technical impact
May affect model selection, inference cost, product capability, and evaluation benchmarks.
[2605.27567] Why LLMs Fail at Causal Discovery and How Interventional Agents Escape
[Submitted on 26 May 2026]
Title:Why LLMs Fail at Causal Discovery and How Interventional Agents Escape
View a PDF of the paper titled Why LLMs Fail at Causal Discovery and How Interventional Agents Escape, by Amartya Roy and 1 other authors
View PDF HTML (experimental)
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
Comments: 9 pages, 3 figures
Subjects:
Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
Cite as: arXiv:2605.27567 [cs.AI]
(or arXiv:2605.27567v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2605.27567
arXiv-issued DOI via DataCite (pending registration)
Submission history
From: Amartya Roy [view email] [v1] Tue, 26 May 2026 18:37:03 UTC (191 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled Why LLMs Fail at Causal Discovery and How Interventional Agents Escape, by Amartya Roy and 1 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.AI
new | recent | 2026-05
Change to browse by:
cs cs.CL
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
Loading...
Data provided by:
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)