Research map / Image generation / Text-to-image diffusion
Achieving Subcategorical Erasure in Text-to-Image Models
Pranav Singh Chib, Pravendra Singh
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 Lipschitz regularization (ablation) | SURE · Guns erasure | Acc_E: 25.20 -> 10.00 | Improved |
| add neutral preservation loss (ablation) | SURE · Guns erasure | Acc_E / Acc_L: Acc_E 30.00 -> 10.00; Acc_L 64.80 -> 67.20 | Improved |
| use 5 perturbation samples instead of 3 (ablation) | SURE · Guns erasure | Acc_E: 30.40 -> 10.00 | Improved |
| increase perturbation samples 5 -> 7 (ablation) | SURE · Guns erasure | Acc_E: 10.00 -> 27.60 | Hurt |
| Gumbel-Softmax instead of Straight-Through Estimator for related concept selection (ablation) | SURE · Guns erasure | Acc_E: 44.80 -> 10.00 | Improved |
| Gumbel-Softmax instead of Sparsemax/Entmax (ablation) | SURE · Guns erasure | Acc_E: 46.80/38.40 -> 10.00 | Improved |
| Gaussian noise sigma 0.1 instead of 0.01 (ablation) | SURE · Guns erasure | Acc_E: 20.00 -> 10.00 | Improved |
| Gaussian noise sigma 0.3 instead of 0.1 (ablation) | SURE · Guns erasure | Acc_E: 10.00 -> 11.20 | No clear effect |
| Lipschitz weight 0.01 instead of 0.1 (ablation) | SURE · Guns erasure | Acc_E: 41.60 -> 10.00 | Improved |
| SURE parent-category erasure vs prior methods | SURE · object categories | harmonic mean H: Guns 76.94 vs ESD 55.11 | Improved |
| SURE vs prior methods on nudity erasure | SURE · I2P | NER-0.3: AGE 5.06 -> 1.70 | Improved |
| SURE vs MACE on locality in style erasure | SURE · artist styles | CLIP score on remaining artists: 31.52 (MACE) vs 30.91 | No clear effect |
Closest papers by meaning
- MACE: Mass Concept Erasure in Diffusion Models (CVPR 2024 · similarity 0.93)
- Localized Concept Erasure for Text-to-Image Diffusion Models Using Training-Free Gated Low-Rank Adaptation (CVPR 2025 · similarity 0.92)
- Fine-Grained Erasure in Text-to-Image Diffusion-based Foundation Models (CVPR 2025 · similarity 0.91)
- Semantic Surgery: Zero-Shot Concept Erasure in Diffusion Models (NeurIPS 2025 · similarity 0.91)
- All but One: Surgical Concept Erasing with Model Preservation in Text-to-Image Diffusion Models (AAAI 2024 · similarity 0.91)
- Precise, Fast, and Low-cost Concept Erasure in Value Space: Orthogonal Complement Matters (CVPR 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.
