Reproduce “ALIVE: Making Streaming Video Models Interactive” 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

Jinzhao Li, Yuan Xu, Jinxuan Wu, Yukun Wang, Yijia Lei, Hanyu Ji, Yinuo Chen, Zining Huang and 8 more

ALIVE trains a video-language model to manage ongoing interactions in streaming video, updating tasks and seeking user input when evidence is insufficient. It combines contrastive supervised fine-tuning with BACA, which adds local behavioral advantages to trajectory-level reinforcement learning. On FSI-Bench’s Free-form Decision Protocol, ALIVE reaches 54.0% accuracy versus 30.1% for GPT-6 Astra, below the 88.5% human reference.

Summary by alphaXiv.

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