AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Preserve and Personalize: Personalized Text-to-Image Diffusion Models without Distributional Drift

Gihoon Kim, Hyungjin Park, Taesup Kim

ICLR 2026 · Image generation › Text-to-image diffusion · approach: Compositional text-to-image guidance · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add Lipschitz-based L2 regularization toward pretrained weights (replacing prior preservation)DreamBooth (SD-1.5) · Personalization benchmark (30 subjects)DINO: 0.6028 -> 0.6394Improved
add Lipschitz-based regularizationDreamBooth (SD-1.5) · Personalization benchmark (30 subjects)CLIP-T: 0.2793 -> 0.2976Improved
add Lipschitz-based regularizationDB-LoRA (SD-1.5) · Personalization benchmark (30 subjects)CLIP-T: 0.3108 -> 0.2975Hurt
add Lipschitz-based regularizationDB-LoRA (SD-1.5) · Personalization benchmark (30 subjects)DINO: 0.5776 -> 0.6038Improved
add Lipschitz-based regularizationCustom Diffusion (SD-1.5) · Personalization benchmark (30 subjects)DINO: 0.5568 -> 0.5638Improved
add Lipschitz-based regularizationCustom Diffusion (SD-XL) · Personalization benchmark (30 subjects)DINO: 0.5337 -> 0.5493Improved
add Lipschitz-based regularizationDB-LoRA (SD-XL) · Personalization benchmark (30 subjects)CLIP-T: 0.3099 -> 0.3014Hurt
add Lipschitz-based regularizationCustom Diffusion (SD-3.0) · Personalization benchmark (30 subjects)CLIP-T: 0.3031 -> 0.2955Hurt
add Lipschitz-based regularizationDB-LoRA (SD-3.0) · Personalization benchmark (30 subjects)CLIP-I: 0.7823 -> 0.7949Improved
increase regularization weight lambda (ablation)Lipschitz-regularized DreamBooth · Personalization benchmark (30 subjects)DINO/CLIP-I vs CLIP-T: higher lambda raises CLIP-T, lower lambda raises DINO and CLIP-IMixed
lower DreamBooth learning rate to 2e-6 instead of using our regularizer (ablation)DreamBooth (SD-1.5) · Personalization benchmark (30 subjects)DINO: 0.6028 -> 0.5707Hurt
add DreamBooth prior-preservation loss on top of Lipschitz regularization (ablation)Lipschitz-regularized DreamBooth · Personalization benchmark (30 subjects)qualityNo clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy