OmniFysics-Nano-V2 Technical Report: Understanding the Physical World Across Modalities
Omni-modal models have expanded multimodal interaction across vision, audio, speech, and language. However, their training is predominantly organized around semantic descriptions and general-purpose objectives, leaving physical attributes, interaction states, and causal mechanisms only partially specified. This gap is...
Yi-Zhou Liu, Jing-Hang Han, Kai Qiu et al.
· 0 citations