Research map / Image generation / Text-to-image diffusion
Multi-subject personalization
Preserve identity and disentangle attributes for personalized multi-subject generation.
82 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2020–2026) · an approach in Text-to-image diffusion, keyword cluster “synthesis · layout · object”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
34 papers in 2023–24 and 47 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 8.7% to 5.5%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2020 · Network-to-Network Translation with Conditional Invertible Neural Networks (NeurIPS 2020 · 10 citations)
- 2023 · Customizable Image Synthesis with Multiple Subjects (NeurIPS 2023 · 4 citations)
- 2024 · PhotoMaker: Customizing Realistic Human Photos via Stacked ID Embedding (CVPR 2024 · 126 citations)
- 2025 · Resolving Multi-Condition Confusion for Finetuning-Free Personalized Image Generation (AAAI 2025 · 11 citations)
- 2026 · Personalize Anything for Free with Diffusion Transformer (AAAI 2026 · 3 citations)
Earliest papers here
- Network-to-Network Translation with Conditional Invertible Neural Networks (NeurIPS 2020 · 10 citations)
- Customizable Image Synthesis with Multiple Subjects (NeurIPS 2023 · 4 citations)
- PhotoMaker: Customizing Realistic Human Photos via Stacked ID Embedding (CVPR 2024 · 126 citations)
What has been tried
65 papers here report 663 measured changes (a component added, removed or swapped, compared with a baseline): helped 74%, no clear effect 4%, hurt 20%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| use stacked ID embedding instead of average embeddings | Embedding composing choices · DINO | 47.0 -> 49.5 | Helped | PhotoMaker: Customizing Realistic Human Photos via Stacked ID Embedding (CVPR 2024 · 126 citations) |
| Remove the HyperNetwork at test time | HyperDreamBooth ablation · DINO | 0.473 -> 0.392 | Hurt | HyperDreamBooth: HyperNetworks for Fast Personalization of Text-to-Image Models (CVPR 2024 · 117 citations) |
| Add expression-aware cross-attention control | Facial expression editing · Expression Coefficients | 0.177 -> 0.193 | Helped | PortraitBooth: A Versatile Portrait Model for Fast Identity-preserved Personalization (CVPR 2024 · 33 citations) |
| increase S3 dataset size from 0.6M to 1.6M samples | DreamBooth test set · DINO | 0.7501 -> 0.7467 | No clear effect | JeDi: Joint-Image Diffusion Models for Finetuning-Free Personalized Text-to-Image Generation (CVPR 2024 · 22 citations) |
| Add 2D hand joint locations to hand embeddings | Test set · FID | 16.839 -> 13.918 | Helped | HanDiffuser: Text-to-Image Generation With Realistic Hand Appearances (CVPR 2024 · 21 citations) |
| Add the middle subject-scene fusion stage (SNF), rather than setting β=α | Person-centric image generation · person-scene coherence | Helped | High-fidelity Person-centric Subject-to-Image Synthesis (CVPR 2024 · 21 citations) | |
| Add meta-learning and EOT transformation sampling | VGGFace2, Trans. Training · SDS | 0.879 -> 0.432 | Helped | MetaCloak: Preventing Unauthorized Subject-driven Text-to-image Diffusion-based Synthesis via Meta-learning (CVPR 2024 · 21 citations) |
| Replace one-hot frame affinity with inter-frame affinity | AnimateBench · image CLIP score | 210.9 -> 225.9 | Helped | PIA: Your Personalized Image Animator via Plug-and-Play Modules in Text-to-Image Models (CVPR 2024 · 21 citations) |
| Replace the CLIP encoder with a face-specific ID encoder | Face identity re-contextualization test set · Face-similarity | 0.266 to 0.302 | Helped | DreamIdentity: Enhanced Editability for Efficient Face-Identity Preserved Image Generation (AAAI 2024 · 14 citations) |
| Add alignment loss to the Stage2 model | DivID-120 · CLIP-T and CLIP-I | 24.91 -> 31.31; 0.624 -> 0.812 | Helped | PuLID: Pure and Lightning ID Customization via Contrastive Alignment (NeurIPS 2024 · 14 citations) |
| Use self-generated samples with automatic filtering in addition to the initial data | Human face domain · CLIP-T | 0.288 -> 0.297 | Helped | Customization Assistant for Text-to-Image Generation (CVPR 2024 · 13 citations) |
| Add layer-wise identifier tokens | ActionBench · action customization quality | Helped | Learning Disentangled Identifiers for Action-Customized Text-to-Image Generation (CVPR 2024 · 12 citations) |
What reviewers will expect
Datasets and baselines named in the 31 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- DreamBench · 4 of 31 papers
- DreamBooth dataset · 3 of 31 papers
- DreamBooth · 3 of 31 papers
- VITON-HD · 2 of 31 papers
- DressCode · 2 of 31 papers
- VGGFace2 · 2 of 31 papers
- CelebA-HQ · 2 of 31 papers
- CustomConcept101 · 2 of 31 papers
Baselines
- IP-Adapter · 10 of 31 papers
- DreamBooth · 9 of 31 papers
- BLIP-Diffusion · 6 of 31 papers
- Textual Inversion · 6 of 31 papers
- Custom Diffusion · 5 of 31 papers
- ELITE · 5 of 31 papers
- PhotoMaker · 4 of 31 papers
- SSR-Encoder · 3 of 31 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 81 papers; from OpenAlex and the papers' affiliations).
- University of Science and Technology of China · 7 papers
- Tsinghua University · 7 papers
- Peking University · 6 papers
- Zhejiang University · 6 papers
- ByteDance · 5 papers
- Alibaba Group · 5 papers
- Google · 4 papers
- Shanghai Jiao Tong University · 3 papers
- Fudan University · 3 papers
- Tencent · 3 papers
Use it
Download BibTeX of 21 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Compositional text-to-image guidance (194 papers)
- Text-driven synthesis and semantic guidance (137 papers)
- Personalized and region-controlled generation (136 papers)
- Layout-to-image control (164 papers)
- Text generation and structured prediction (65 papers)
Elsewhere in the topic
- T2I evaluation and controllable adapters (149 papers)
- Inference-time T2I improvement (117 papers)
- Preference optimization for diffusion (82 papers)
- Concept erasure in diffusion models (76 papers)
- Multi-concept composition (56 papers)
- Fairness and bias in generative models (52 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).