Identity-GRPO: Optimizing Multi-Human Identity-preserving Video Generation via Reinforcement Learning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911408081338368 |
|---|---|
| author | Meng, Xiangyu Zhang, Zixian Zhang, Zhenghao Liao, Junchao Qin, Long Wang, Weizhi |
| author_facet | Meng, Xiangyu Zhang, Zixian Zhang, Zhenghao Liao, Junchao Qin, Long Wang, Weizhi |
| contents | While advanced methods like VACE and Phantom have advanced video generation for specific subjects in diverse scenarios, they struggle with multi-human identity preservation in dynamic interactions, where consistent identities across multiple characters are critical. To address this, we propose Identity-GRPO, a human feedback-driven optimization pipeline for refining multi-human identity-preserving video generation. First, we construct a video reward model trained on a large-scale preference dataset containing human-annotated and synthetic distortion data, with pairwise annotations focused on maintaining human consistency throughout the video. We then employ a GRPO variant tailored for multi-human consistency, which greatly enhances both VACE and Phantom. Through extensive ablation studies, we evaluate the impact of annotation quality and design choices on policy optimization. Experiments show that Identity-GRPO achieves up to 18.9% improvement in human consistency metrics over baseline methods, offering actionable insights for aligning reinforcement learning with personalized video generation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_14256 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Identity-GRPO: Optimizing Multi-Human Identity-preserving Video Generation via Reinforcement Learning Meng, Xiangyu Zhang, Zixian Zhang, Zhenghao Liao, Junchao Qin, Long Wang, Weizhi Computer Vision and Pattern Recognition While advanced methods like VACE and Phantom have advanced video generation for specific subjects in diverse scenarios, they struggle with multi-human identity preservation in dynamic interactions, where consistent identities across multiple characters are critical. To address this, we propose Identity-GRPO, a human feedback-driven optimization pipeline for refining multi-human identity-preserving video generation. First, we construct a video reward model trained on a large-scale preference dataset containing human-annotated and synthetic distortion data, with pairwise annotations focused on maintaining human consistency throughout the video. We then employ a GRPO variant tailored for multi-human consistency, which greatly enhances both VACE and Phantom. Through extensive ablation studies, we evaluate the impact of annotation quality and design choices on policy optimization. Experiments show that Identity-GRPO achieves up to 18.9% improvement in human consistency metrics over baseline methods, offering actionable insights for aligning reinforcement learning with personalized video generation. |
| title | Identity-GRPO: Optimizing Multi-Human Identity-preserving Video Generation via Reinforcement Learning |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2510.14256 |