Research map / Image generation / Text-to-image diffusion
ConceptPrism: Concept Disentanglement in Personalized Diffusion Models via Residual Token Optimization
Minseo Kim, Minchan Kwon, Dongyeun Lee, Yunho Jeon, Junmo Kim
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 image-wise residual tokens (ablation) | ConceptPrism token optimization · DreamBench | CLIP-T: 0.345 -> 0.358 | Improved |
| add image-wise residual tokens (ablation) | ConceptPrism token optimization · DreamBench | DINO: 0.197 -> 0.183 | Hurt |
| add exclusion loss (ablation) | ConceptPrism token optimization · DreamBench | DINO: 0.183 -> 0.207 | Improved |
| add exclusion loss (ablation) | ConceptPrism token optimization · DreamBench | CLIP-T: 0.358 -> 0.354 | No clear effect |
| add residual-token initialization from image descriptions (ablation) | ConceptPrism token optimization · DreamBench | DINO: 0.207 -> 0.210 | Improved |
| freeze residual tokens at initial description instead of optimizing (ablation) | ConceptPrism · DreamBench | CLIP-T: 0.357 -> 0.313 | Hurt |
| shorten initial residual description from long to short (ablation) | ConceptPrism · DreamBench | CLIP-T: 0.357 -> 0.353 | No clear effect |
| add residual tokens (n>=1) versus none (ablation) | ConceptPrism · DreamBench | DINO: 0.197 -> 0.209 | Improved |
| increase residual token length from 1 to 8 (ablation) | ConceptPrism · DreamBench | CLIP-T: 0.354 -> 0.357 | No clear effect |
| use ConceptPrism instead of DreamBooth-LoRA | ConceptPrism · DreamBench | DINO: 0.182 -> 0.210 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- DisenBooth — DisenBooth: Identity-Preserving Disentangled Tuning for Subject-Driven Text-to-Image Generation (ICLR 2024) · better on 3 of 3 shared results
- ELITE — ELITE: Encoding Visual Concepts into Textual Embeddings for Customized Text-to-Image Generation (ICCV 2023) · better on 3 of 3 shared results
- DisEnvisioner — DisEnvisioner: Disentangled and Enriched Visual Prompt for Customized Image Generation (ICLR 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- The Hidden Language of Diffusion Models (ICLR 2024 · similarity 0.90)
- Concept Weaver: Enabling Multi-Concept Fusion in Text-to-Image Models (CVPR 2024 · similarity 0.90)
- Attention Calibration for Disentangled Text-to-Image Personalization (CVPR 2024 · similarity 0.89)
- ConceptBed: Evaluating Concept Learning Abilities of Text-to-Image Diffusion Models (AAAI 2024 · similarity 0.89)
- You Can’t Have It Both Ways: Concept Entanglement Limits Diffusion Model Unlearning (NeurIPS 2026 · similarity 0.89)
- Personalized Residuals for Concept-Driven Text-to-Image Generation (CVPR 2024 · similarity 0.89)
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.
