Accident-Bench/

Dataset

AccidentBench: Benchmarking Multimodal Understanding and Reasoning in Vehicle Accidents and Beyond

{ "id": int, "dataset": "str", // e.g., sub dataset filename "scene_name": "str", // e.g., video filename "reasoning_style": "str", // e.g., temporal_reasoning, intent_goal_reasoning, etc. "question": "str", // The reasoning question related to the scene "ground_truth": "str", // Correct… See the full description on the dataset page: https://huggingface.co/datasets/Accident-Bench/Dataset.