Research map / Video / Video generation
Efficient Training for Human Video Generation with Entropy-Guided Prioritized Progressive Learning
Changlin Li, Jiawei Zhang, Shuhao Liu, Sihao Lin, Zeyi Shi, Zhihui Li, Xiaojun Chang
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 CEI training priority (ablation) | Ent-Prog · TikTok video | FID-VID: 37.43 -> 32.15 | Improved |
| add CEI training priority (ablation) | Ent-Prog · TikTok video | FVD: 285.94 -> 264.03 | Improved |
| add adaptive progressive schedule (ablation) | Ent-Prog · TikTok video | FVD: 382.41 -> 264.03 | Improved |
| add adaptive progressive schedule (ablation) | Ent-Prog · TikTok video | FID-VID: 44,31 -> 32.15 | Improved |
| add adaptive progressive schedule (ablation) | Ent-Prog · TikTok video | Training speedup: 2.11x -> 1.52x | Hurt |
| add CEI training priority (ablation) | Ent-Prog · TikTok video | Training speedup: 1.95x -> 1.52x | Hurt |
| Ent-Prog instead of full training | Ent-Prog · Bilibili video (200k steps) | FVD: 132.06 -> 119.77 | Improved |
| Ent-Prog instead of full training | Ent-Prog · TikTok video | FVD: 385.64 -> 264.03 | Improved |
| Ent-Prog instead of full training | Ent-Prog · UBC-Fashion video | FVD: 79.94 -> 79.76 | No clear effect |
| Ent-Prog instead of full training | Ent-Prog · Bilibili video (100k steps) | L1: 1.31e-05 -> 1.43e-05 | Hurt |
| Ent-Prog instead of full training | Ent-Prog · Bilibili video (200k steps) | Memory (GB): 72 -> 53 | Improved |
| Ent-Prog instead of full training | Ent-Prog · TikTok image | FID: 61.29 -> 59.18 | Improved |
Closest papers by meaning
- Training-Free Efficient Video Generation via Dynamic Token Carving (NeurIPS 2025 · similarity 0.87)
- AR-Diffusion: Asynchronous Video Generation with Auto-Regressive Diffusion (CVPR 2025 · similarity 0.87)
- FreeNoise: Tuning-Free Longer Video Diffusion via Noise Rescheduling (ICLR 2024 · similarity 0.87)
- Sparse Video-Gen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsity (ICML 2025 · similarity 0.87)
- SaRA: High-Efficient Diffusion Model Fine-tuning with Progressive Sparse Low-Rank Adaptation (ICLR 2025 · similarity 0.86)
- From Slow Bidirectional to Fast Autoregressive Video Diffusion Models (CVPR 2025 · similarity 0.86)
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.
