1
VisPhyWorld: Probing Physical Reasoning via Code-Driven Video Reconstruction
重新定义物理推理评估：代码驱动视频重建让MLLMs必须真正理解物理规律，而非表面识别。
arXiv:2602.13294v3 Announce Type: replace Abstract: Evaluating whether Multimodal Large Language Models (MLLMs) genuinely reason about physical dynami…