Research map / Image generation / Text-to-image diffusion
MOSAIC: Multi-Subject Personalized Generation via Correspondence-Aware Alignment and Disentanglement
Dong She, Siming Fu, Mushui Liu, Qiaoqiao Jin, Hualiang Wang, Mu Liu, Jidong Jiang
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 semantic correspondence attention loss L_SCA (ablation) | MOSAIC (FLUX.1-dev + LoRA) · DreamBench (multi-subject) | CLIP-I: 73.45 -> 75.89 | Improved |
| add semantic correspondence attention loss L_SCA (ablation) | MOSAIC (FLUX.1-dev + LoRA) · DreamBench (multi-subject) | DINO: 52.03 -> 55.99 | Improved |
| add semantic correspondence attention loss L_SCA (ablation) | MOSAIC (FLUX.1-dev + LoRA) · DreamBench (multi-subject) | CLIP-T: 29.90 -> 31.10 | Improved |
| add multi-reference disentanglement loss L_MD on top of L_SCA (ablation) | MOSAIC (FLUX.1-dev + LoRA) · DreamBench (multi-subject) | CLIP-T: 31.10 -> 32.40 | Improved |
| add multi-reference disentanglement loss L_MD on top of L_SCA (ablation) | MOSAIC (FLUX.1-dev + LoRA) · DreamBench (multi-subject) | DINO: 55.99 -> 56.83 | Improved |
| add multi-reference disentanglement loss L_MD on top of L_SCA (ablation) | MOSAIC (FLUX.1-dev + LoRA) · DreamBench (multi-subject) | CLIP-I: 75.89 -> 76.30 | Improved |
| MOSAIC vs best baseline (XVerse) on overall score | MOSAIC · XVerseBench | Overall score: 73.40 -> 76.04 | Improved |
| MOSAIC vs DreamO on DPG (single-subject) | MOSAIC · XVerseBench (single-subject) | DPG: 96.93 -> 96.55 | No clear effect |
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 5 of 5 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
- Mind-the-Glitch: Visual Correspondence for Detecting Inconsistencies in Subject-Driven Generation (NeurIPS 2025 · similarity 0.86)
- ARGUS: Stacked Multi-View Identity Mosaic Injection for Subject-Preserving Video Generation (NeurIPS 2026 · similarity 0.86)
- DynASyn: Multi-Subject Personalization Enabling Dynamic Action Synthesis (AAAI 2025 · similarity 0.85)
- Selectively Informative Description can Reduce Undesired Embedding Entanglements in Text-to-Image Personalization (CVPR 2024 · similarity 0.85)
- Refine-by-Align: Reference-Guided Artifacts Refinement through Semantic Alignment (ICLR 2025 · similarity 0.85)
- DisenBooth: Identity-Preserving Disentangled Tuning for Subject-Driven Text-to-Image Generation (ICLR 2024 · 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.
