Em-Garde: A Propose-Match Framework for Proactive Streaming Video Understanding

Yikai Zheng, Xin Ding, Yifan Yang, Shiqi Jiang, Hao Wu, Qianxi Zhang, Weijun Wang, Ting Cao, Yunxin Liu · 2026

Read Em-Garde: A Propose-Match Framework for Proactive Streaming Video Understanding by Yikai Zheng, Xin Ding, Yifan Yang, Shiqi Jiang, Hao Wu, Qianxi Zhang, Weijun Wang, Ting Cao, Yunxin Liu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-19054.txt.

Description

Zheng and co-authors build a system for video models that answer before being asked again. In streaming video, users pose a query once and expect a timely response when the moment comes. Checking each frame for a trigger is slow or error prone. Their Em-Garde splits the task in two. At query time, a parser turns the request into clear visual targets. During the stream, a light matcher scans for those targets and fires replies. Tests on two benches show gains in accuracy and speed.

Catalog record

CREATORYikai Zheng, Xin Ding, Yifan Yang, Shiqi Jiang, Hao Wu, Qianxi Zhang, Weijun Wang, Ting Cao, Yunxin Liu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE49727
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.19054

Browse the full library · About the Tower of Babel