Jinzhao Li et al., Tsinghua University · Interactive Streaming Video

ALIVE: Making Streaming Video Models Interactive

ALIVE is an interactive streaming video model that can proactively monitor video, coordinate streaming tasks and seek feedback from users. The authors build two training sets, ALIVE-SFT-288K and ALIVE-RL-3K, and train in two stages: contrastive supervised fine-tuning, then reinforcement learning with Behavior-Aligned Credit Assignment (BACA). They also introduce FSI-Bench for evaluating feedback seeking and task completion. The paper reports the highest overall scores among compared models on four public benchmarks and FSI-Bench. Licence and availability are not stated on the page.

Open the source ↗

Similar Video & Animation releases

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