VideoEvoHarness constructs structured memory records describing entities and events in local video segments.
Video agents improve their own harnesses using unlabeled video
VideoEvoHarness jointly adapts task generation and agent workflows while keeping the underlying vision-language models frozen.
Chinese Tech
Weixin Xu · Yifan Liu · Yi Wang · Mingfei Rong · Zhenyu Yang · Zhisheng Zhong · +2 more
Institute of Automation, Chinese Academy of Sciences · University of Chinese Academy of Sciences · Jiangsu University of Science and Technology · AICOMO · Inner Mongolia University
Research Digest··3 min read
Xu and colleagues present a framework that turns unlabeled videos into supervision for improving video-understanding agents.
Why this paper
From Institute of Automation, Chinese Academy of Sciences and 5 others
In one line
VideoEvoHarness evolves video agent harnesses from unlabeled videos without annotations or parameter updates, improving accuracy.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§