arXiv · 2609.31675
Active Causal Discovery Benchmark: Evaluating LLM Agents Under Budgeted Interventions
Abstract
We introduce the Active Causal Discovery Benchmark (ACDB), an SCM-grounded environment for evaluating whether LLM agents recover causal graph structure from observations and budget-constrained hard interventions. ACDB pairs a linear-Gaussian world generator with a fixed observe-intervene-submit API and a three-layer scoring contract that separates skeleton recovery, DAG recovery, and intervention efficiency. On the current six-level ladder, PC with a greedy active orientation heuristic is the strongest non-oracle method (directed F1 42.7%, SHD 4.79), ahead of Claude Sonnet 4.6 raw active (31.7%, 7.25) and GPT-5.4 raw active (22.9%, 9.27). The most informative diagnostic is the precision-recall decomposition: PC under-commits with high precision, LLMs over-commit with lower precision, and statistical-tool access often increases abstention rather than useful intervention. A structure-blind random DAG baseline reaches 23.6% directed F1 on this dense v0 ladder; a density probe lowers this floor to 16.9%, motivating the v1 calibration pass. The current results should therefore be read as a benchmark audit and calibration report, not as evidence that current LLMs solve active causal discovery.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Sagar Deb, Devam Shah, Ashwanth Krishnan. 2026-09-16. Active Causal Discovery Benchmark: Evaluating LLM Agents Under Budgeted Interventions. https://searcharxiv.com/abs/2609.31675
Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Searcharxiv by Indiciae: BibTeX · RIS.