An Efficient Streaming Video Understanding Framework with Agentic Control

Benchmark Model Rank Results
video-question-answering-on-ovo-benchR3-Streaming (3B / 4B-Thinking)–Real-Time Visual Perception: 63.58Backward Tracing: 50.23…
video-question-answering-on-ovo-benchR3-Streaming (3B / 7B-Instruct)–Real-Time Visual Perception: 65.12Backward Tracing: 44.64…
video-question-answering-on-ovo-benchR3-Streaming (7B / 4B-Thinking)–Real-Time Visual Perception: 71.89Backward Tracing: 51.27…