The authors treat compressed video as a sequence of state updates rather than a collection of separately encoded frames.
Carried codec states improve video models’ temporal reasoning
RESUME accumulates motion and residual updates in a recurrent latent state, preserving event order without increasing predictive-frame token budgets.
Chinese Tech
Can Zhang · Xiaotian Han · Junyuan Shang · Yuchen Ding · Zhenyu Zhang · Shuohuan Wang · +2 more
Beijing University of Chemical Technology · Baidu, Inc.
Research Digest··3 min read
Zhang et al.
Why this paper
From Baidu, Inc. and Beijing University of Chemical Technology
In one line
RESUME improves temporal video reasoning by updating a latent state with motion and residual signals from codec predictions.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ·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.
§