Research map / Video / Video generation
DCARL: A Divide-and-Conquer Framework for Autoregressive Long-Trajectory Video Generation
Junyi Ouyang, Wenbin Teng, Gonglin Chen, Yajie Zhao, Haiwei Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add image-level keyframe encoding instead of temporal compression (ablation) | DCARL keyframe model · ODV-YT 16s keyframes | FID: 19.6 -> 16.3 | Improved |
| add image-level keyframe encoding instead of temporal compression (ablation) | DCARL keyframe model · ODV-YT 16s keyframes | ATE: 0.155 -> 0.100 | Improved |
| increase keyframe generation stride from 4 to 16 (interpolation stride 16) (ablation) | DCARL · ODV-YT 16s | FID: 21.4 -> 18.8 | Improved |
| use generation and interpolation stride 8/8 instead of 4/4 (ablation) | DCARL · ODV-YT 16s | ATE: 0.148 -> 0.096 | Improved |
| use stride 16/16 instead of 8/8 (ablation) | DCARL · ODV-YT 16s | ATE: 0.096 -> 0.104 | Hurt |
| add noisy keyframes in both training and inference (ablation) | DCARL interpolation · ODV-YT 16s | FVD: 246.5 -> 201.7 | Improved |
| add noisy keyframes in both training and inference (ablation) | DCARL interpolation · ODV-YT 16s | FID: 19.0 -> 19.3 | No clear effect |
| add noisy keyframes in training only (ablation) | DCARL interpolation · ODV-YT 16s | FVD: 246.5 -> 240.9 | No clear effect |
| add latent substitution only to overlap (ablation) | DCARL segment continuity · Segment boundary | PSNR: 18.15 -> 17.95 | Hurt |
| add training of the continuity mechanism on top of latent substitution (ablation) | DCARL segment continuity · Segment boundary | PSNR: 17.95 -> 18.75 | Improved |
| add keyframe conditioning in interpolation (ablation) | DCARL · ODV-YT 32s | Overall FVD: 376.7 -> 203.7 | Improved |
| add keyframe conditioning in interpolation (ablation) | DCARL · ODV-YT 32s | ATE: 0.387 -> 0.237 | Improved |
Closest papers by meaning
- InfinityStar: Unified Spacetime AutoRegressive Modeling for Visual Generation (NeurIPS 2025 · similarity 0.89)
- ARLON: Boosting Diffusion Transformers with Autoregressive Models for Long Video Generation (ICLR 2025 · similarity 0.89)
- VideoMAR: Autoregressive Video Generation with Continuous Tokens (NeurIPS 2025 · similarity 0.88)
- Generative Pre-trained Autoregressive Diffusion Transformer (NeurIPS 2025 · similarity 0.88)
- Ca2-VDM: Efficient Autoregressive Video Diffusion Model with Causal Generation and Cache Sharing (ICML 2025 · similarity 0.88)
- From Slow Bidirectional to Fast Autoregressive Video Diffusion Models (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
