Shuhan Li, Huankang Guan, Xintong Guo, Yang Bo, Yuyang He, Dunyuan Xu, Jingyan Chen, Fang Liu · Streaming Video LLM

Beyond Answering: Learning to Speak First and Act in Streaming Video (AURA 2.0)

AURA 2.0 is a streaming video large language model that can start conversations without a user query (Model-Initiated Conversation), call external tools based on the visual stream (Stream-Aware Tool Calling), and ground objects marked by users to frame-aligned bounding boxes. The authors also add long-horizon interaction memory and train on a dedicated streaming interaction dataset using a three-stage scheme. They report leading performance among open-source streaming models on public streaming benchmarks, but the page does not state a licence or release of weights.

Open the source ↗

Similar Video & Animation releases

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.