Benchmark specifically designed to diagnose hallucinations at different reasoning stages in medical multimodal language models.
Large-scale multimodal dataset with 100K samples for audio-visual reasoning with structured scripts and evidence chains. 19 upvotes on HuggingFace.