dorsal/arxiv
View SchemaStep-by-Step Causality: Transparent Causal Discovery with Multi-Agent Tree-Query and Adversarial Confidence Estimation
| Authors | Ziyi Ding, Chenfei Ye-Hao, Zheyuan Wang, Xiao-Ping Zhang |
|---|---|
| Categories | |
| ArXiv ID | 2601.10137vv1 |
| URL | https://arxiv.org/abs/2601.10137 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
Causal discovery aims to recover ``what causes what'', but classical constraint-based methods (e.g., PC, FCI) suffer from error propagation, and recent LLM-based causal oracles often behave as opaque, confidence-free black boxes. This paper introduces Tree-Query, a tree-structured, multi-expert LLM framework that reduces pairwise causal discovery to a short sequence of queries about backdoor paths, (in)dependence, latent confounding, and causal direction, yielding interpretable judgments with robustness-aware confidence scores. Theoretical guarantees are provided for asymptotic identifiability of four pairwise relations. On data-free benchmarks derived from Mooij et al. and UCI causal graphs, Tree-Query improves structural metrics over direct LLM baselines, and a diet--weight case study illustrates confounder screening and stable, high-confidence causal conclusions. Tree-Query thus offers a principled way to obtain data-free causal priors from LLMs that can complement downstream data-driven causal discovery. Code is available at https://anonymous.4open.science/r/Repo-9B3E-4F96.
{
"annotation_id": "48df8ea3-5480-4d43-9ae3-97c02fb135b4",
"date_created": "2026-02-17T05:53:23.557000Z",
"date_modified": "2026-02-17T05:53:23.557000Z",
"file_hash": "3c5dee2683a4b3dace953350c7bc5cec226c0a13808c31f2371cab6474b242b1",
"private": false,
"record": {
"abstract": "Causal discovery aims to recover ``what causes what\u0027\u0027, but classical constraint-based methods (e.g., PC, FCI) suffer from error propagation, and recent LLM-based causal oracles often behave as opaque, confidence-free black boxes. This paper introduces Tree-Query, a tree-structured, multi-expert LLM framework that reduces pairwise causal discovery to a short sequence of queries about backdoor paths, (in)dependence, latent confounding, and causal direction, yielding interpretable judgments with robustness-aware confidence scores. Theoretical guarantees are provided for asymptotic identifiability of four pairwise relations. On data-free benchmarks derived from Mooij et al. and UCI causal graphs, Tree-Query improves structural metrics over direct LLM baselines, and a diet--weight case study illustrates confounder screening and stable, high-confidence causal conclusions. Tree-Query thus offers a principled way to obtain data-free causal priors from LLMs that can complement downstream data-driven causal discovery. Code is available at https://anonymous.4open.science/r/Repo-9B3E-4F96.",
"arxiv_id": "2601.10137",
"authors": [
"Ziyi Ding",
"Chenfei Ye-Hao",
"Zheyuan Wang",
"Xiao-Ping Zhang"
],
"categories": [
"cs.LG",
"cs.AI",
"stat.ML"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "Step-by-Step Causality: Transparent Causal Discovery with Multi-Agent Tree-Query and Adversarial Confidence Estimation",
"url": "https://arxiv.org/abs/2601.10137",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "a86ce58d-8ecd-49d6-8669-2e5cdb3a7c38",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}