Research map / Video / Video generation
FloodDiffusion: Tailored Diffusion Forcing for Streaming Motion Generation
Yiyi Cai, Yuhan Wu, Kunhang Li, You Zhou, Bo Zheng, Haiyang Liu
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 |
|---|---|---|---|
| use bi-directional attention instead of causal attention (ablation) | FloodDiffusion · HumanML3D | FID: 3.377 -> 0.057 | Improved |
| use bi-directional attention instead of causal attention (ablation) | FloodDiffusion · HumanML3D | R@3: 0.625 -> 0.810 | Improved |
| use lower-triangular timestep scheduler instead of random scheduler (ablation) | FloodDiffusion · HumanML3D | FID: 3.883 -> 0.057 | Improved |
| use lower-triangular scheduler instead of random (ablation) | FloodDiffusion · HumanML3D | R@3: 0.532 -> 0.810 | Improved |
| use classifier-free guidance scale 6 instead of no guidance (ablation) | FloodDiffusion · HumanML3D | FID / MM-Dist: FID 0.057, MM-Dist 2.852 at CFG=6 | Improved |
| swap Wan CausalVAE for MotionStreamer CausalVAE (ablation) | FloodDiffusion · Causal VAE reconstruction | FID: 0.0290 -> 0.0271 | No clear effect |
| use Wan CausalVAE instead of VQ-VAE (ablation) | FloodDiffusion · Causal VAE reconstruction | FID: 0.1259 -> 0.0290 | Improved |
| diffusion forcing framework vs MotionStreamer on streaming | FloodDiffusion · BABEL streaming | AUJ: 16.57 -> 14.05 | Improved |
| diffusion forcing vs MoMask on FID | FloodDiffusion · HumanML3D | FID: 0.045 -> 0.057 | Mixed |
Closest papers by meaning
- AMD: Autoregressive Motion Diffusion (AAAI 2024 · similarity 0.88)
- StreamingT2V: Consistent, Dynamic, and Extendable Long Video Generation from Text (CVPR 2025 · similarity 0.87)
- Rethinking Diffusion for Text-Driven Human Motion Generation: Redundant Representations, Evaluation, and Masked Autoregression (CVPR 2025 · similarity 0.87)
- AR-Diffusion: Asynchronous Video Generation with Auto-Regressive Diffusion (CVPR 2025 · similarity 0.87)
- LLM-grounded Video Diffusion Models (ICLR 2024 · similarity 0.87)
- Human Motion Diffusion as a Generative Prior (ICLR 2024 · 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.
