KlingAI Research (Yuqi Tang et al.) · Egocentric Video Captioner

EgoCaptioner

EgoCaptioner is a captioning approach for egocentric video that pairs a video-level overview with timed, adaptively detailed steps and marked key moments. It is trained on 50,000 examples built with a coarse-to-fine annotation pipeline across eight egocentric video sources, and refined with GRPO using factual, coverage, temporal-alignment and anti-fragmentation rewards. The authors report leading results among evaluated open-source models across five egocentric benchmarks, and say it surpasses Gemini 3.1 Pro on PhysCapBench. Licence and code 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.