Tag
This paper proposes a validation framework for using Large Language Models to extract causal relations from social media posts during disasters. It evaluates the effectiveness of LLMs in identifying cause-effect relationships and compares them against expert-grounded reference graphs to assess reliability and risks.