Research map / Image generation / Text-to-image diffusion
Forget Many, Forget Right: Scalable and Precise Concept Unlearning in Diffusion Models
Kaiyuan Deng, Gen Li, Yang Xiao, Bo Hui, Xiaolong Ma
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 |
|---|---|---|---|
| use ScaPre closed-form unlearning instead of RECE | SD v1.5 Imagenette unlearning · Imagenette (10 classes) | avg unlearn accuracy (%): 4.9 -> 0.8 | Improved |
| use ScaPre instead of UCE/RECE for 50-concept unlearning | SD v1.5 · ImageNet-Diversi50 (50 classes) | CLIP score (COCO): UCE 22.23 -> ScaPre 29.41 | Improved |
| use ScaPre instead of UCE for precise unlearning | SD v1.5 · ImageNet-Confuse5 | preserve accuracy (%): 5.6 -> 76.3 | Improved |
| use ScaPre instead of UCE | SD v1.5 · ImageNet-Confuse5 | unlearn accuracy (%): 2.9 -> 5.8 | Hurt |
| use ScaPre instead of MACE for generation quality | SD v1.5 · Imagenette (10 classes) | CLIP score (COCO): 31.02 -> 30.43 | Hurt |
| use ScaPre for explicit-content removal | SD v1.4 · I2P NudeNet total detections | total detections: 743 -> 54 | Improved |
| use ScaPre against white-box UnlearnDiffAtk | unlearned SD · Robustness: UnlearnDiffAtk | attack success rate (%): 67.1 -> 60.9 | Improved |
Closest papers by meaning
- You Can’t Have It Both Ways: Concept Entanglement Limits Diffusion Model Unlearning (NeurIPS 2026 · similarity 0.91)
- Unlearning Concepts in Diffusion Model via Concept Domain Correction and Concept Preserving Gradient (AAAI 2025 · similarity 0.91)
- Boosting Alignment for Post-Unlearning Text-to-Image Generative Models (NeurIPS 2024 · similarity 0.90)
- CURE: Concept Unlearning via Orthogonal Representation Editing in Diffusion Models (NeurIPS 2025 · similarity 0.89)
- Fine-Grained Erasure in Text-to-Image Diffusion-based Foundation Models (CVPR 2025 · similarity 0.89)
- CASL: Concept-Aligned Sparse Latents for Interpreting Diffusion Models (NeurIPS 2026 · 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.
