Existing post-training methods usually reduce a generated video to one reward value.
Local and global rewards improve long-horizon video consistency
LoGo assigns spatially targeted feedback to 3D errors while retaining whole-video rewards for camera control and visual quality.
Academic
Ziqi Ma · Shreya Sharma · Mohamed El Banani · Katja Schwarz · Chongjie Ye · Chao-Yuan Wu · +5 more
California Institute of Technology · World Labs
Research Digest··2 min read
Ma et al.
Why this paper
From California Institute of Technology and World Labs
In one line
Local-global rewards improve 3D consistency in long-horizon camera-controlled video generation.
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 (2 noted)
- ·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.
Concepts
§