Research map / Image generation / Fast diffusion sampling
Unlearning in Diffusion Models: A Unified Framework with KL Divergence and Likelihood Constraints
Shervin Khalafi, Alejandro Ribeiro, Dongsheng Ding
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 constraint thresholds (reverse KL) instead of equal-weight penalties (ablation) | reverse KL unlearning · Stable Diffusion v1.4 multi-concept unlearning | KL to pretrained model at same unlearning level | Improved |
| use constraint thresholds (reverse KL) instead of equal-weight penalties (ablation) | reverse KL unlearning · Stable Diffusion v1.4 multi-concept unlearning | max CLIP similarity to unlearned concepts | Improved |
| learn per-sample dual weights (forward KL constraints) instead of equal weights (ablation) | forward KL unlearning · CelebA-HQ DDPM, 3 memorized images | KID at same max SSCD | Improved |
| use likelihood constraint instead of concept erasing (ablation) | likelihood-constrained unlearning · Three-Gaussian mixture | KL to retained modes at same unlearning likelihood | Improved |
| use likelihood constraint instead of concept erasing (ablation) | likelihood-constrained unlearning · Stable Diffusion v1.4, unlearn horse keep cowboy | KID and KL to pretrained model | Improved |
| use likelihood constraint instead of reverse KL constraint | likelihood-constrained unlearning · Three-Gaussian mixture | retained modes preserved | Improved |
Closest papers by meaning
- Data Unlearning in Diffusion Models (ICLR 2025 · similarity 0.92)
- Constrained Diffusion Models via Dual Training (NeurIPS 2024 · similarity 0.91)
- Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models (NeurIPS 2026 · similarity 0.91)
- Training Data Attribution in Diffusion Models via Mirrored Unlearning and Noise-Consistent Skew (NeurIPS 2026 · similarity 0.91)
- The Illusion of Unlearning: The Unstable Nature of Machine Unlearning in Text-to-Image Diffusion Models (CVPR 2025 · similarity 0.90)
- The Emergence of Reproducibility and Consistency in Diffusion Models (ICML 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.
