Research map / Image generation / Text-to-image diffusion
Meta-Unlearning on Diffusion Models: Preventing Relearning Unlearned Concepts
Hongcheng Gao, Tianyu Pang, Chao Du, Taihang Hu, Zhijie Deng, Min Lin
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 meta objective to unlearning | ESD-u-1 · FT on HRM-m 300 steps | nudity score (%): 34.51 -> 26.06 | Improved |
| add meta objective to unlearning | ESD-f-3 · FT on HRM-m 300 steps | nudity score (%): 55.63 -> 33.10 | Improved |
| add meta objective to unlearning | RECE · FT on HRM-m 300 steps | nudity score (%): 22.54 -> 13.38 | Improved |
| add meta objective to unlearning | SDD · FT on HRM-m 300 steps | nudity score (%): 54.23 -> 48.59 | Improved |
| add meta objective to unlearning | UCE · FT on HRM-s 300 steps | nudity score (%): 57.75 -> 42.96 | Improved |
| add meta objective to unlearning | UCE · FT on CLEAN 300 steps | nudity score (%): 26.76 -> 4.23 | Improved |
| add meta objective to unlearning | ESD-u-3 · COCO 10K FID before FT | FID: 20.52 -> 19.98 | No clear effect |
| add meta objective to unlearning | UCE · COCO 10K FID before FT | FID: 17.59 -> 19.20 | Hurt |
| add meta objective to unlearning | ESD-f-3 · COCO 10K FID before FT | FID: 21.38 -> 18.54 | Improved |
| add meta objective to unlearning | RECE · UnlearnDiffAtk | attack success rate (%): 64.79 -> 62.68 | No clear effect |
| add meta objective to unlearning | ESD-f-3 · NSFW unsafe score, FT on HRM-m 50 steps | unsafe score: 39.44 -> 8.45 | Improved |
| increase inner finetuning steps M from 1 to 5 (ablation) | Meta-unlearned ESD-u-3 · nudity after FT | nudity score (%): 26.76 -> 32.39 | Hurt |
Closest papers by meaning
- The Illusion of Unlearning: The Unstable Nature of Machine Unlearning in Text-to-Image Diffusion Models (CVPR 2025 · similarity 0.94)
- UnlearnCanvas: Stylized Image Dataset for Enhanced Machine Unlearning Evaluation in Diffusion Models (NeurIPS 2024 · similarity 0.91)
- Towards Resilient Safety-driven Unlearning for Diffusion Models against Downstream Fine-tuning (NeurIPS 2025 · similarity 0.91)
- Data Unlearning in Diffusion Models (ICLR 2025 · similarity 0.91)
- Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models (NeurIPS 2026 · similarity 0.90)
- Score Forgetting Distillation: A Swift, Data-Free Method for Machine Unlearning in Diffusion Models (ICLR 2025 · 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.
