Binxu Li, Haoyi Duan, Yuhui Zhang, Yaohui Zhang, Zihao Lin, Kaituo Feng, Suozhi Huang, Xiangyi Li, Yu Li, Chunyuan Li, Shilong Liu, Mengdi Wang
VideoGen-Agent is a multimodal agent that uses reinforcement learning to coordinate tools for video generation, improving accuracy in complex prompts involving identity, physics, and temporal structure. It achieves significant gains on a new benchmark by combining tool use with iterative refinement and quality evaluation.
The approach enables more reliable and controllable video generation for real-world applications requiring precise visual storytelling, such as education, simulation, and content creation. It demonstrates that agents can effectively leverage external tools to surpass standalone models, even as underlying generation models improve.
Thank you to arXiv for use of its open access interoperability. arxiv.org. aibrief.ai is not affiliated with or endorsed by arXiv.