Open-source medical video understanding system uAI-NEXUS-MedVLM released

让大模型理解真实医疗视频，全球首个开源技术方案来了！

United Imaging Intelligence released uAI-NEXUS-MedVLM for medical video understanding, with a CVPR 2026 paper. MedVidBench has 532k video-instruction pairs across 8 medical sources and 8 tasks. Qwen2.5-VL-7B SFT reached 89.4% CVS accuracy; GPT-5.4 scored 16.4%.

Why it matters: HKR-H/K/R all pass: the story has a real-medical-video open-source hook, concrete 530K+ data scale, 8 tasks, and a 89.4% vs 16.4% result. The medical focus keeps it in the 78–84 band.