Reproduce “WeMM-Video: Pushing the Frontiers of Open-Source Omni-Modal Video Understanding” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Changli Tang, Jie Yang, Tianyi Wang, Lei Li, Daoze Zhang, Zijie Xin, Zhan Liu, Li Li and 4 more

WeMM-Video combines video frames and audio in chronological order for a language model to answer questions requiring cross-modal and temporal reasoning. It offers direct instruct and explicit thinking modes. On Video-MME with audio, its 35B-A3B thinking model scores 86.4, below Gemini-3.0-Flash at 88.6. The reported benchmark results support performance on tested tasks, not universal gains or elimination of hallucinations.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.