Research map / Image generation / Text-to-image diffusion
Receler: Reliable Concept Erasing of Text-to-Image Diffusion Models via Lightweight Erasers
Chi-Pin Huang, Kai-Po Chang, Chung-Ting Tsai, Yung-Hsuan Lai, Fu-En Yang, Yu-Chiang Frank Wang
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 |
|---|---|---|---|
| replace full-model fine-tuning with a lightweight eraser (ablation) | Receler · CIFAR-10 object erasure | Acc_R / Acc_L: Acc_R 39.4 -> 34.9; Acc_L 75.2 -> 77.4 | Improved |
| add concept-localized regularization to the eraser (ablation) | Receler · CIFAR-10 object erasure | Acc_L: 77.4 -> 79.8 | Improved |
| add adversarial prompt learning to the eraser (ablation) | Receler · CIFAR-10 object erasure | Acc_R: 34.9 -> 20.9 | Improved |
| add adversarial prompt learning without regularization (ablation) | Receler · CIFAR-10 object erasure | Acc_L: 77.4 -> 76.2 | Hurt |
| combine regularization and adversarial prompt learning (ablation) | Receler · CIFAR-10 object erasure | Acc_R / Acc_L: Acc_R 17.6; Acc_L 82.1 | Improved |
| use Receler instead of ESD | SD v1.4 concept erasing · CIFAR-10 paraphrased prompts | Acc_R: 39.9 -> 17.6 | Improved |
| use Receler instead of ESD/UCE on I2P | SD v1.4 safety erasing · I2P overall | inappropriate proportion (%): 31.3 -> 27.0 (vs UCE) | Improved |
| use Receler on COCO-30K versus FMN | nudity-erased SD v1.4 · COCO-30K | FID-30K: 14.10 vs 13.52 (FMN) | No clear effect |
| use Receler instead of other erasers against Ring-A-Bell nudity prompts | nudity-erased SD v1.4 · Ring-A-Bell nudity | failure rate: 49.5% -> 1.1% | Improved |
Later papers that compared with this method
4 later accepted papers report Receler as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- CASteer: Cross-Attention Steering for Controllable Concept Erasure (ICLR 2026 · better on 2 of 3 shared results)
- CE-SDWV: Effective and Efficient Concept Erasure for Text-to-Image Diffusion Models via a Semantic-Driven Word Vocabulary (IJCV 2026 · better on 1 of 1 shared result)
- Semantic Surgery: Zero-Shot Concept Erasure in Diffusion Models (NeurIPS 2025 · better on 2 of 2 shared results)
- SAeUron: Interpretable Concept Unlearning in Diffusion Models with Sparse Autoencoders (ICML 2025 · better on 0 of 1 shared result)
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
- Concept Pinpoint Eraser for Text-to-image Diffusion Models via Residual Attention Gate (ICLR 2025 · similarity 0.92)
- Localized Concept Erasure for Text-to-Image Diffusion Models Using Training-Free Gated Low-Rank Adaptation (CVPR 2025 · similarity 0.92)
- EraseAnything: Enabling Concept Erasure in Rectified Flow Transformers (ICML 2025 · similarity 0.90)
- Erasing Undesirable Concepts in Diffusion Models with Adversarial Preservation (NeurIPS 2024 · similarity 0.90)
- All but One: Surgical Concept Erasing with Model Preservation in Text-to-Image Diffusion Models (AAAI 2024 · similarity 0.90)
- When Are Concepts Erased From Diffusion Models? (NeurIPS 2025 · similarity 0.90)
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.
