Research map / Image generation / Text-to-image diffusion
Zero-Shot Image Personalization from Personas
Harini S I, Somesh Singh, Yaman K Singla, David Doermann, Rajiv Shah
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 persona conditioning to LLM prompt rewriting | 14 LLMs · ZIP-Bench | ClipScore: gains 3% (Qwen3-30B-A3B) to 20% (Claude-Sonnet-4) | Improved |
| Add persona conditioning (zero-shot) | GPT-4o · ZIP-Bench | PIGReward: 65.3 -> 73.4 | Improved |
| Use persona zero-shot instead of fine-tuned DrUM | DrUM · benchmarks (3 of 4) | CLIPScore: e.g. MovieLens 72.8 -> 76.4 | Improved |
| Use persona zero-shot instead of fine-tuned DrUM | DrUM · PIP | CLIPScore: 66.9 -> 60.2 | Hurt |
| Use ZIPPY 3-shot instead of TV 3-shot | TV · ZIP-Bench | CLIPScore: 62.4 -> 66.7 | Improved |
| Increase ZIPPY shots from 3 to 5 | ZIPPY · ZIP-Bench | PIGReward: 77.5 -> 82.8 | Improved |
| Use persona conditioning instead of fine-tuning for pluralism | ViPer/DrUM · all users | CMMD: 0.55 / 0.31 -> 0.16 | Improved |
| Use persona conditioning for demographic alignment (IPF-normalized) | DrUM · RapidData demographics | PIGReward: 56.8 -> 73.0 | Improved |
| Use MoE LLMs instead of dense Qwen3 | Qwen3-30B-A3B / 235B-A22B · ZIP-Bench | ClipScore | Hurt |
Closest papers by meaning
- Visual Persona: Foundation Model for Full-Body Human Customization (CVPR 2025 · similarity 0.88)
- JeDi: Joint-Image Diffusion Models for Finetuning-Free Personalized Text-to-Image Generation (CVPR 2024 · similarity 0.86)
- DynASyn: Multi-Subject Personalization Enabling Dynamic Action Synthesis (AAAI 2025 · similarity 0.85)
- Personalized Preference Fine-tuning of Diffusion Models (CVPR 2025 · similarity 0.85)
- MoLE: Enhancing Human-centric Text-to-image Diffusion via Mixture of Low-rank Experts (NeurIPS 2024 · similarity 0.85)
- PersonaHOI: Effortlessly Improving Face Personalization in Human-Object Interaction Generation (CVPR 2025 · 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.
