Research map / Video / Video generation
Implicit Preference Alignment for Human Image Animation
Yuanzhi Wang, Xuhua Ren, Jiaxiang Cheng, bing ma, Kai Yu, Tianxiang Zheng, Qinglin Lu, Zhen Cui
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 |
|---|---|---|---|
| Align VACE-14B with IPA on self-generated good samples (ablation) | VACE-14B · TikTok | FVD: 427 -> 288 | Improved |
| Add IPA to VACE-14B on the hand-dynamics benchmark (ablation) | VACE-14B · Hand-dynamics benchmark | FID-VID: 12.5 -> 8.1 | Improved |
| Add Hand-Aware Local Optimization on top of IPA (ablation) | IPA · TikTok | FID-VID: 7.9 -> 5.9 | Improved |
| Add Hand-Aware Local Optimization on top of IPA (ablation) | IPA · Hand-dynamics benchmark | SSIM: 0.741 -> 0.757 | Improved |
| Use small beta (200) in the implicit-reward loss (ablation) | IPA · TikTok and hand benchmark | FID-VID, FVD, SSIM, PSNR | Hurt |
| Use large beta (above 600) in the implicit-reward loss (ablation) | IPA · TikTok and hand benchmark | FID-VID, FVD, SSIM, PSNR | Hurt |
| Raise hand weighting lambda from 0.1 to 10 (ablation) | IPA with HALO · TikTok | FID-VID: 7.4 -> 5.9 | Improved |
| Raise hand weighting lambda from 10 to 100 (ablation) | IPA with HALO · TikTok | FID-VID: 5.9 -> 6.8 | Hurt |
| Use IPA-aligned VACE-14B instead of Wan-Animate | IPA · TikTok | FVD: 316 -> 255 | Improved |
| Use IPA-aligned VACE-14B instead of Wan-Animate on hand regions | IPA · Hand regions | PSNR-Hand: 14.1 -> 18.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MimicMotion — MimicMotion: High-Quality Human Motion Video Generation with Confidence-aware Pose Guidance (ICML 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Direct Post-Training Preference Alignment for Multi-Agent Motion Generation Model Using Implicit Feedback from Pre-training Demonstrations (ICLR 2025 · similarity 0.87)
- HandEdit: A Unified Benchmark for Egocentric Human-to-Robot Dexterous Hand Image Editing (NeurIPS 2026 · similarity 0.86)
- PrefPaint: Aligning Image Inpainting Diffusion Model with Human Preference (NeurIPS 2024 · similarity 0.86)
- FoundHand: Large-Scale Domain-Specific Learning for Controllable Hand Image Generation (CVPR 2025 · similarity 0.86)
- OHTA: One-shot Hand Avatar via Data-driven Implicit Priors (CVPR 2024 · similarity 0.86)
- Re-HOLD: Video Hand Object Interaction Reenactment via adaptive Layout-instructed Diffusion Model (CVPR 2025 · similarity 0.85)
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.
