Research map / Image generation / Text-to-image diffusion
Mitigating Semantic Collapse in Generative Personalization with Test-Time Embedding Adjustment
Anh Tuan Bui, Thuy-Trang Vu, Trung Le, Junae Kim, Tamas Abraham, Rollin Omari, Amardeep Kaur, Dinh Phung
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 test-time embedding adjustment (TEA) | DreamBooth · CC101 | CLIP-T full prompt: +1.87 | Improved |
| add TEA | Textual Inversion · CC101 | CLIP-T full prompt: +0.64 | Improved |
| add TEA | Textual Inversion · CC101 | CLIP-I: -2.34 | Hurt |
| add TEA | DreamBooth · CelebA | CLIP-I: -2.37 | Hurt |
| add TEA | DreamBooth · CC101 | CLIP-I: +4.57 | Improved |
| add TEA | Custom Diffusion · CelebA | DINO-I: +3.09 | Improved |
| add TEA | Custom Diffusion / ClassDiffusion · CC101 | CLIP-T prompt-part: -0.13 / -0.12 | No clear effect |
| add TEA | EasyControl · CC101 Pet Dog | DINO-I: 43.71 -> 48.32 | Improved |
| add TEA | OminiControl · Subject Penguin | VLM-P: 86.25 -> 90.50 | Improved |
| add TEA | ReVersion · Relationship Carved by | CLIP-T full prompt: 27.74 -> 30.17 | Improved |
| increase rotation factor alpha (ablation) | TEA · CC101 DreamBooth | CLIP-T full prompt: 26.5 -> 28.2 | Improved |
| increase rotation factor alpha (ablation) | TEA · CC101 DreamBooth | CLIP-I: 73.5 -> 71.1 | Hurt |
Closest papers by meaning
- Salient Concept-Aware Generative Data Augmentation (NeurIPS 2025 · similarity 0.89)
- AttnDreamBooth: Towards Text-Aligned Personalized Text-to-Image Generation (NeurIPS 2024 · similarity 0.88)
- CoRe: Context-Regularized Text Embedding Learning for Text-to-Image Personalization (AAAI 2025 · similarity 0.88)
- ClassDiffusion: More Aligned Personalization Tuning with Explicit Class Guidance (ICLR 2025 · similarity 0.88)
- Tailored Visions: Enhancing Text-to-Image Generation with Personalized Prompt Rewriting (CVPR 2024 · similarity 0.88)
- 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.
