Research map / Image generation / Text-to-image diffusion
CASteer: Cross-Attention Steering for Controllable Concept Erasure
Tatiana Gaintseva, Andreea-Maria Oncescu, Chengcheng Ma, Ziquan Liu, Martin Benning, Gregory Slabaugh, Jiankang Deng, Ismail Elezi
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 intermediate clipping (ablation) | CASteer · I2P nudity removal | total detected nude parts: 12 -> 7 | Improved |
| add intermediate clipping (ablation) | CASteer · I2P inappropriate content | overall inappropriate proportion (%): 28.94 -> 25.58 | Improved |
| add intermediate clipping (ablation) | CASteer · Remove Van Gogh | Acc_e: 0.35 -> 0.25 | Improved |
| add intermediate clipping (ablation) | CASteer · Remove Van Gogh | LPIPS_e: 0.46 -> 0.44 | Hurt |
| add intermediate clipping (ablation) | CASteer · Remove Kelly McKernan | Acc_e: 0.05 -> 0.05 | No clear effect |
| steer cross-attention outputs rather than other layer types (ablation) | CASteer · DiT backbone layers | concept erasure effectiveness | Improved |
| use 50 prompt pairs for steering vector (ablation) | CASteer · concrete concept erasure | concept capture | No clear effect |
| use CASteer instead of SD v1.4 unmodified | SD v1.4 · I2P nudity | total detected nude parts: 646 -> 7 | Improved |
| use CASteer (clip) instead of Receler | Receler · I2P inappropriate content | overall inappropriate proportion (%): 27.0 -> 25.58 | Improved |
| use CASteer instead of SPM for implicit Mickey removal | SPM · user study | images preferred (%): 2.17% -> 84.32% | Improved |
| use CASteer instead of SPM for Snoopy removal | SPM · user study | images containing Snoopy (%): 44.4 -> 3.3 | Improved |
| use CASteer instead of SPM for preserving Mickey when removing Snoopy | SPM · user study | average rank (lower = more Mickey kept): 1.05 -> 2.22 | Hurt |
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.
- SAFREE — SAFREE: Training-Free and Adaptive Guard for Safe Text-to-Image And Video Generation (ICLR 2025) · better on 4 of 4 shared results
- Receler — Receler: Reliable Concept Erasing of Text-to-Image Diffusion Models via Lightweight Erasers (ECCV 2024) · better on 2 of 3 shared results
- SAeUron — SAeUron: Interpretable Concept Unlearning in Diffusion Models with Sparse Autoencoders (ICML 2025) · better on 1 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 0.86)
- Enhancing Creative Generation on Stable Diffusion-based Models (CVPR 2025 · similarity 0.86)
- The Hidden Language of Diffusion Models (ICLR 2024 · similarity 0.86)
- Towards Safe Concept Transfer of Multi-Modal Diffusion via Causal Representation Editing (NeurIPS 2024 · similarity 0.86)
- Unlearning Concepts in Diffusion Model via Concept Domain Correction and Concept Preserving Gradient (AAAI 2025 · similarity 0.86)
- Concept Replacer: Replacing Sensitive Concepts in Diffusion Models via Precision Localization (CVPR 2025 · similarity 0.86)
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.
