一作ACL 2026Poster· Natural Language Processing
N oisy C ausal: A Benchmark for Evaluating Causal Reasoning Under Structured Noise
Zhi Xu · Yun Fu
摘要
Causal reasoning in natural language requires identifying relevant variables, understanding their interactions, and reasoning about effects and interventions, often under noisy or ambiguous conditions. While large language models (LLMs) exhibit strong general reasoning abilities, they struggle to...