Reproduce “ViDecide: Teaching Video Diffusion Models to Reason and Plan Step by Step” 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
Kaiwen Zhang, Yiming Gao, Zeqi Xiao, Xintao Wang, Pengfei Wan, Kun Gai, Xingang Pan
The authors introduce ViDecide, a video diffusion method that generates task states one at a time and signals completion with a stop frame. It emphasizes high-noise decisions and uses Corrective Fine-tuning to target failures. Across four benchmark columns, it averages 83.8% Strict Success Rate, versus 23.3% for fine-tuning and 67.8% for S2PD; results include simulation tasks, not real-world robot control.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.