Research map / Image generation / Text-to-image diffusion
Guiding Diffusion Models with Adaptive Negative Sampling Without External Resources
Alakh Desai, Nuno Vasconcelos
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 adaptive negative sampling (ANSWER) to CFG | SDXL · ImageNet / AE / Pick-a-Pic / DrawBench / PartiPrompts | win rate (HPSv2): 41-52% | Improved |
| add ANSWER sampling | SDXL · AMT human evaluation | preference rate: 46-61% vs 8-20% | Improved |
| add ANSWER sampling | SD and SDXL · AE, Pick-a-Pic, DrawBench, PartiPrompts | Inception Score | Hurt |
| increase number of DNS steps K up to a moderate value (ablation) | SDXL + ANSWER · ImageNet | FID and HPSv2 | Improved |
| increase K too high (e.g. K=15) (ablation) | SDXL + ANSWER · ImageNet | image quality, FID and HPSv2: K=15 | Hurt |
| add mean and variance normalization of negative noise (ablation) | SDXL + ANSWER · qualitative example | image sharpness | Improved |
Closest papers by meaning
- Dynamic Negative Guidance of Diffusion Models (ICLR 2025 · similarity 0.92)
- AP-Adapter: Improving Generalization of Automatic Prompts on Unseen Text-to-Image Diffusion Models (NeurIPS 2024 · similarity 0.90)
- Normalized Attention Guidance: Universal Negative Guidance for Diffusion Models (NeurIPS 2025 · similarity 0.90)
- On Discrete Prompt Optimization for Diffusion Models (ICML 2024 · similarity 0.90)
- Prompting Hard or Hardly Prompting: Prompt Inversion for Text-to-Image Diffusion Models (CVPR 2024 · similarity 0.89)
- Learning to Sample Effective and Diverse Prompts for Text-to-Image Generation (CVPR 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.
