Research map / Image generation / Text-to-image diffusion
ViPer: Visual Personalization of Generative Models via Individual Preference Learning
Sogand Salehi, Mahdi Shafiei, Roman Bachmann, Teresa Yeo, Amir Zamir
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 |
|---|---|---|---|
| personalize with ViPer visual preferences vs. no personalization | Stable Diffusion XL · User study (20 users, 8 images per prompt) | top-1 accuracy (%): 12.5 -> 86.1 | Improved |
| personalize with ViPer vs. other users' personalizations | Stable Diffusion XL · User study (20 users) | top-1 accuracy (%): 12.5 -> 65.4 | Improved |
| inject visual preference embeddings plus guidance instead of LLM prompt rewriting | ViPer · 1000 agents, personalized vs non-personalized | accuracy (%), proxy M: 83.9 -> 89.1 | Improved |
| ViPer instead of Prompt Personalization for prompt fidelity | ViPer · Prompt-image alignment, 1000 agents | accuracy (%), CLIP: 83.5 -> 90.3 | Improved |
| use positive preferences only vs negative only for identifying agents (ablation) | GPT-4 VP vs. VP test · Identifiability test | top-1 accuracy (%): 45.8 -> 92.4 | Improved |
| increase number of comments used to extract visual preferences (ablation) | ViPer VPE · qualitative single-user example | alignment with user preference | Improved |
| fine-tune Stable Diffusion on the proxy reward instead of embedding conditioning | Stable Diffusion · qualitative | image quality | Hurt |
| increase beta from 0 to 1 (ablation) | ViPer · qualitative | degree of personalization | Improved |
Closest papers by meaning
- Multimodal Large Language Models Make Text-to-Image Generative Models Align Better (NeurIPS 2024 · similarity 0.89)
- Tailored Visions: Enhancing Text-to-Image Generation with Personalized Prompt Rewriting (CVPR 2024 · similarity 0.88)
- Personalized Preference Fine-tuning of Diffusion Models (CVPR 2025 · similarity 0.88)
- Preference Adaptive and Sequential Text-to-Image Generation (ICML 2025 · similarity 0.87)
- A User-Friendly Framework for Generating Model-Preferred Prompts in Text-to-Image Synthesis (AAAI 2024 · similarity 0.87)
- Yo'Chameleon: Personalized Vision and Language Generation (CVPR 2025 · similarity 0.87)
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.
