Modelpedia
Work in progress
About
Findings
Models
Concepts
Methods
Datasets
Sources
Light
Dark
CAS: A Probability-Based Approach for Universal Condition Alignment Score
2024-01-16
· ICLR 2024 spotlight ·
anchor
Findings
IC-783
Stable Diffusion v1.5's conditional probability pθ(x|c) is heavily biased by the unconditional probability pθ(x), making it unreliable as a condition-alignment metric
IC-784
Pre-trained scoring models (CLIP Score, HPS, Image Reward, Pick Score) underperform on domain-specific fine-tuned diffusion models
IC-785
Different released diffusion models produce images with distinguishable probability signatures, enabling source attribution
IC-786
CLIP-ViT (LC) achieves 0.87 accuracy and 0.91 average precision on fake image detection