Researchers have created a dataset called ConstructCIE to help extract causal information from narratives about construction accidents. The dataset includes manually annotated OSHA reports and uses a hierarchical schema to categorize accident types, causes, and supporting evidence. Experiments showed that while some models can predict accident types and identify broad causes, they struggle with precise extraction of specific details. More accurate domain grounding and evidence extraction are needed for reliable causal information extraction. --- Why it matters: This dataset matters because it provides a valuable resource for researchers working on natural language processing (NLP) tasks related to construction safety and accident analysis. By improving the accuracy of causal information extraction, developers can create more effective tools for identifying potential hazards and preventing accidents in the construction industry. Source: https://arxiv.org/abs/2608.06495