AI Paper Atlas

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

ICML 2026 · Image generation › Fast diffusion sampling · approach: Guidance for conditional diffusion · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
use CCFG instead of naive negative prompting for class negation with a positive classImageNet-1k class-conditional model · ImageNet-1k fine-grained pairsoverall accuracy: 0.4957 -> 0.9039Improved
use CCFG instead of DNG for negationImageNet-1k class-conditional model · ImageNet-1k fine-grained pairsoverall accuracy: 0.8212 -> 0.9039Improved
use CCFG instead of NPImageNet-1k class-conditional model · ImageNet-1k fine-grained pairsFD-DINOv2: 456.78 -> 190.71Improved
use positive CCFG instead of CFGMNIST and CIFAR10 class guidance · MNIST, CIFAR10classification accuracy vs FIDNo clear effect
use negative CCFG instead of NP and DNGMNIST and CIFAR10 class removal · MNIST, CIFAR10removal accuracy vs FIDImproved
use CCFG instead of CFG for positive promptsStableDiffusion 1.5 · GenEval overalloverall score: 0.407 -> 0.432Improved
use CCFG instead of CFG for positive promptsSDXL · DPGBench overalloverall score: 0.751 -> 0.752No clear effect
use CCFG instead of NP with positive and negative promptsStableDiffusion 1.5 · PIE-Bench-NegVLM overall: 0.7303 -> 0.7777Improved
use CCFG instead of Perp-Neg and SLD on SDXLSDXL · PIE-Bench-NegVLM overall: 0.8128 (SLD) vs 0.8120 (CCFG)No clear effect
use constant temperature tau instead of timestep-dependent tau (ablation)CCFG · generalsample stabilityNo clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy