Reproduce “AskStream: When and What to Ask Proactively under Uncertainty in Streaming Video Understanding” 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
Lu Wang, Zhuoran Jin, Longxiang Wang, Junhong Chen, Yupu Hao, Yao Xu, Daojian Zeng, Yubo Chen and 3 more
AskStream evaluates streaming video assistants that can wait, ask users for missing information, or answer. The benchmark covers uncertainty in video evidence and user input, and uses timing-aware Ask-F1 to score appropriate asking and subsequent answering. In multiple-choice evaluation, Gemini-3.8 Flash achieves 52.18% overall Ask-F1, the best score in that setting; its open-ended score is 29.19%.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.