Research map / Image generation / Text-to-image diffusion
GrOCE : Graph-Guided Online Concept Erasure for Text-to-Image Diffusion Models
Ning Han, Zhenyu Ge, Feng Han, Yuhua Sun, Chengqing Li, Jingjing Chen
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 Identify (adaptive cluster identification) before Sever (ablation) | GrOCE · Erase Snoopy | non-target FID: 426.74 -> 0 | Improved |
| add Identify before Sever (ablation) | GrOCE · Erase Snoopy | target CS: 14.51 -> 16.92 | Hurt |
| use graph-guided cluster erasure instead of prior inference-time methods (AdaVD) | GrOCE vs AdaVD · Erase Snoopy, Mickey & Spongebob | target CS (Snoopy): 19.17 -> 16.92 | Improved |
| use graph-guided erasure instead of fine-tuning (MACE) | GrOCE vs MACE · Erase Snoopy | non-target FID (Mickey): 98.42 -> 0 | Improved |
| use graph-guided erasure instead of AdaVD for art style erasure | GrOCE vs AdaVD · Erase Van Gogh | target CS: 24.26 -> 23.28 | Improved |
| training-free precomputed graph instead of fine-tuning or data preparation | GrOCE · 10-concept erasure runtime | total time (s): 3.59 (AdaVD) -> 1.73; 8529.01 (ConAbl) -> 1.73 | Improved |
| increase cluster size K (ablation) | GrOCE · Erase Snoopy hyperparameter study | non-target FID | Hurt |
| lower projection threshold delta (ablation) | GrOCE · Erase Snoopy hyperparameter study | non-target preservation | Hurt |
| apply GrOCE to SDXL, FLUX.1-schnell and other SD versions | GrOCE · Snoopy, Mickey, Spongebob, Pikachu prompts | erasure and preservation | No clear effect |
Closest papers by meaning
- MACE: Mass Concept Erasure in Diffusion Models (CVPR 2024 · similarity 0.91)
- Semantic Surgery: Zero-Shot Concept Erasure in Diffusion Models (NeurIPS 2025 · similarity 0.91)
- Localized Concept Erasure for Text-to-Image Diffusion Models Using Training-Free Gated Low-Rank Adaptation (CVPR 2025 · similarity 0.90)
- Fantastic Targets for Concept Erasure in Diffusion Models and Where To Find Them (ICLR 2025 · similarity 0.90)
- Consistency-Preserving Concept Erasure via Unsafe–Safe Pairing and Directional Fisher-weighted Adaptation (NeurIPS 2026 · similarity 0.89)
- Fine-Grained Erasure in Text-to-Image Diffusion-based Foundation Models (CVPR 2025 · similarity 0.88)
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.
