🧪 Test?View on arXiv
PEEK: One-Step-Look-Ahead Exposure Bias Correction for Diffusion Sampling
Dacheng Guo, Author 2, Author 3, Author 4, Author 5
diffusion modelsimage generationexposure biassampling techniques
2611.05407
Builder Relevance
2h ago80%
Abstract
PEEK proposes a novel training-free deterministic sampler plugin that mitigates exposure bias in diffusion image generation.
Reality Card
Core Claim
PEEK improves image generation quality by reducing exposure bias, achieving FID improvements up to 49.3% with minimal extra computation cost.
Method / Result
FID improvements up to 49.3% at minimal extra computation cost.
Limitations
The one-time calibration process may vary in time based on model complexity, which could affect reproducibility.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.
No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.