Research map / Image generation / Autoregressive and latent generation
Scaling Multi-Identity Consistency for Image Customization via Multi-to-Multi Matching Paradigm
Yufeng Cheng, Wenxu Wu, Shaojin Wu, Mengqi Huang, Fei Ding, Qian He
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 ReReFL with SIR reward on top of plain SFT (ablation) | UNO · XVerseBench Multi-Subject | ID-Sim: 33.94 -> 65.16 | Improved |
| plain SFT on the multi-identity data (ablation) | UNO · XVerseBench Multi-Subject | ID-Sim: 31.82 -> 33.94 | No clear effect |
| use MIMR instead of SIR reward (ablation) | UMO on UNO · XVerseBench Multi-Subject | ID-Conf: 65.28 -> 78.06 | Improved |
| use MIMR instead of SIR reward (ablation) | UMO on UNO · XVerseBench Multi-Subject | ID-Sim: 65.16 -> 69.09 | Improved |
| apply UMO training | UNO · XVerseBench Single-Subject | ID-Sim: 47.91 -> 80.89 | Improved |
| apply UMO training | OmniGen2 · XVerseBench Single-Subject | ID-Sim: 62.41 -> 91.57 | Improved |
| apply UMO training | OmniGen2 · XVerseBench Multi-Subject | ID-Conf: 62.02 -> 77.74 | Improved |
| apply UMO training | OmniGen2 · OmniContext | Overall: 7.18 -> 7.16 | No clear effect |
| apply UMO training | UNO · XVerseBench Single-Subject | IP-Sim: 80.40 -> 77.09 | Hurt |
| apply UMO training | UNO · OmniContext | ID-Conf: 4.91 -> 6.60 | 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.
- XVerse — XVerse: Consistent Multi-Subject Control of Identity and Semantic Attributes via DiT Modulation (NeurIPS 2025) · better on 3 of 3 shared results
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
- RealisID: Scale-Robust and Fine-Controllable Identity Customization via Local and Global Complementation (AAAI 2025 · similarity 0.87)
- ID-Patch: Robust ID Association for Group Photo Personalization (CVPR 2025 · similarity 0.85)
- Illuminating Visual Identity in Universal Multimodal Embeddings (CVPR 2026 · similarity 0.84)
- DreamSalon: A Staged Diffusion Framework for Preserving Identity-Context in Editable Face Generation (CVPR 2024 · similarity 0.84)
- FaceA-Net: Facial Attribute-Driven ID Preserving Image Generation Network (AAAI 2025 · similarity 0.84)
- HiFi-Portrait: Zero-shot Identity-preserved Portrait Generation with High-fidelity Multi-face Fusion (CVPR 2025 · similarity 0.84)
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.
