Can MiniMax-H3 Reason About the Physical World? An Evaluation of Omni-Modal Generative Model
Dataset Analysis
Evaluates MiniMax-H3's physical world reasoning via omni-modal tasks combining text, frames, audio, and video; achieves 41.97% success across 517 instances.
Provenance
Collected from papers.
Derived from paper: Can MiniMax-H3 Reason About the Physical World? An Evaluation of Omni-Modal Generative Model