Reproduce “VideoGen-Agent: Reinforcing Video Generation Agents” 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

Binxu Li, Haoyi Duan, Yuhui Zhang, Yaohui Zhang, Zihao Lin, Kaituo Feng, Hui Yuan, Suozhi Huang and 5 more

VideoGen-Agent coordinates retrieval, simulation, generation, and verification for prompts involving identity, motion, and event order. A policy selects tools and adapts its action to results, and is trained with supervised fine-tuning followed by reinforcement learning. On VABench, it scores 75.6 with Toolset 1 versus 56.5 for its base generator; Toolset 2 scores 86.1, with the policy fixed and generation tools changed.

Summary by alphaXiv.

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