CausaLab: A Scalable Environment for Interactive Causal Discovery Toward AI Scientists

Junlin Yang, Dylan Zhang, Xiangchen Song, Qirun Dai, Xiao Liu, Yuen Chen, Aniket Vashishtha, Jing Shi, Chenhao Tan, Hao Peng

arXiv:2605.26029 · 2026-05-30 공개 · arXiv · PDF

llm-agents causal-discovery causal-reasoning intervention-strategy causal-graph-recovery resonance-frequency synthetic-laboratory consistency-verification

Abstract

We introduce CausaLab, a scalable environment for evaluating interactive causal discovery by LLM agents. Unlike prior evaluations, CausaLab evaluates both whether an agent can solve a problem using causal evidence and whether its answer is grounded in a faithful recovered causal mechanism. Each episode places an agent in a synthetic laboratory: it receives prior measurement records, intervenes on a manipulator crystal, and predicts the resonance frequency of a held-out reactor crystal governed by the same mechanism. The hidden data-generating process is a randomly sampled structural causal model (SCM), so success requires recovering both a causal graph and structural equations rather than recalling prior knowledge. Experiments show a persistent gap between prediction and mechanism recovery: in the purely observational 6-node setting, GPT-5.2-high reaches 92% task accuracy but only 0.471 all-edge F_1. Mixed observation-intervention strategies improve structural fidelity, while pure intervention remains difficult even for strong agents. We identify premature stopping as a major weakness and show that consistency verification mitigates it. CausaLab therefore separates predictive success from causal understanding and exposes current LLM agents' limits as experimental causal reasoners.

한국어 요약

📋 한 줄 요약

**[Causal Discovery / LLM Agent]** CausaLab이 hidden SCM이 지배하는 합성 lab에 LLM agent를 배치해 예측·메커니즘 회복 양면 평가 — GPT-5.2-high가 6-node 관측만에서 task 92%·all-edge F1 0.471로 큰 gap, premature stopping이 주 약점.

🎯 핵심 기여도

💡 핵심 아이디어

LLM agent의 causal discovery 평가는 정답 예측만이 아니라 faithful causal mechanism 회복 여부도 함께 측정해야 하며, 합성 SCM 기반 lab 환경에서 두 지표를 분리 평가하면 LLM agent가 experimental causal reasoner로서 가진 한계가 드러난다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: LLM agent의 causal reasoning을 단순 정답에서 mechanism 회복으로 확장, predictive success ≠ causal understanding의 gap 정량 노출, premature stopping 등 구체적 실패 mode 진단, scalable 환경으로 agent 발전 추적 가능. **한계**: 합성 SCM 환경과 실세계 causal discovery의 격차, 6-node 중심 결과로 scaling 추가 검증, crystal·resonance 시나리오의 도메인 특화.

🚀 실용적 활용