Research map / Image generation / Fast diffusion sampling
ContrastiveCFG: Guiding Diffusion Sampling by Contrasting Positive and Negative Concepts
Jinho Chang, Changsun Lee, Hyungjin Chung, Jong Chul Ye
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 |
|---|---|---|---|
| use CCFG instead of naive negative prompting for class negation with a positive class | ImageNet-1k class-conditional model · ImageNet-1k fine-grained pairs | overall accuracy: 0.4957 -> 0.9039 | Improved |
| use CCFG instead of DNG for negation | ImageNet-1k class-conditional model · ImageNet-1k fine-grained pairs | overall accuracy: 0.8212 -> 0.9039 | Improved |
| use CCFG instead of NP | ImageNet-1k class-conditional model · ImageNet-1k fine-grained pairs | FD-DINOv2: 456.78 -> 190.71 | Improved |
| use positive CCFG instead of CFG | MNIST and CIFAR10 class guidance · MNIST, CIFAR10 | classification accuracy vs FID | No clear effect |
| use negative CCFG instead of NP and DNG | MNIST and CIFAR10 class removal · MNIST, CIFAR10 | removal accuracy vs FID | Improved |
| use CCFG instead of CFG for positive prompts | StableDiffusion 1.5 · GenEval overall | overall score: 0.407 -> 0.432 | Improved |
| use CCFG instead of CFG for positive prompts | SDXL · DPGBench overall | overall score: 0.751 -> 0.752 | No clear effect |
| use CCFG instead of NP with positive and negative prompts | StableDiffusion 1.5 · PIE-Bench-Neg | VLM overall: 0.7303 -> 0.7777 | Improved |
| use CCFG instead of Perp-Neg and SLD on SDXL | SDXL · PIE-Bench-Neg | VLM overall: 0.8128 (SLD) vs 0.8120 (CCFG) | No clear effect |
| use constant temperature tau instead of timestep-dependent tau (ablation) | CCFG · general | sample stability | No clear effect |
Closest papers by meaning
- Rectified Diffusion Guidance for Conditional Generation (CVPR 2025 · similarity 0.93)
- CFG++: Manifold-constrained Classifier Free Guidance for Diffusion Models (ICLR 2025 · similarity 0.91)
- Adaptive Guidance: Training-free Acceleration of Conditional Diffusion Models (AAAI 2025 · similarity 0.90)
- No Training, No Problem: Rethinking Classifier-Free Guidance for Diffusion Models (ICLR 2025 · similarity 0.90)
- TCFG: Tangential Damping Classifier-free Guidance (CVPR 2025 · similarity 0.90)
- Eliminating Oversaturation and Artifacts of High Guidance Scales in Diffusion Models (ICLR 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.
