Reproduce “StreamVibe: Proactive Video Agents for Evolving User Intent” 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
Chengzhi Liu, Yuxiang Lai, Yuzhe Yang, Yanqing Liu, Jiheng Hou, Yue Cao, Haonan Wang, Ruixi Fan and 8 more
StreamVibe pairs a streaming policy that decides when to speak with a Harness tracking active requests, task progress, and prior responses. In full-stream evaluation on Dynamic Intents, adding the Harness to the same StreamVibe-VL (RL) policy raises Content F1 from 5.25 to 8.91 and lowers near-miss FPR from 17.11 to 5.68 at ±2 seconds, supporting improvement in this setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.