MCG-NJU/OneStreamer

OneStreamer Unifying Perception, Memory, and Proactive Responsein Streaming Video Interaction Xiangyu Zeng1,, Yuandong Yang1,, Zhiqiu Zhang2,4,, Yuhan Zhu1,, Xinhao Li1,, Qingyi Si3, Dingyu Yao6, Changlian Ma1,2, Haoran Chen1,2, Xinyu Chen1,2, Yansong Shi2,5, Junhao Zhou6 Yifei Li9, Jun Zhang1,3, Chuanyu Qin6, Chenxu Yang6, Xinlei Yu7, Kun Ouyang8 Yuchen Shao6, Qianshan Wei6, Changhai Zhou10, Jun Gao11, Jiaqi Wang3, Limin Wang1,† 1NJU · 2PJLAB · 3JD · 4SJTU · 5USTC · 6CAS 7CUHK · 8PKU · 9THU · 10FDU · 11ZJU Overview · Method · Models and data · Results · Quick start · Citation Perceive the present. Remember the past. Respond at the right time. OneStreamer is a streaming video LLM that jointly learns to record evidence and respond through a shared proactive generation process. It preserves time-grounded factual memory before future questions are known, interprets incoming visual evidence, and decides when enough evidence is available to respond. Built on Qwen3-VL-4B-Instruct, OneStreamer-4B…