anchor
Findings
- IC-1337InstructPix2Pix is more effective at editing color than at preserving category in visual concept editing [eval]
- IC-331In Stable Diffusion v1.5, specific tokens in text prompts drive memorization, and GPT-4-based perturbation of high-attribution tokens reduces SSIM similarity to training images while maintaining CLIP score [eval]
- IC-498Multimodal foundation models exhibit severe group unfairness, with race and age biases more pronounced than gender bias in text-to-image models while gender bias is stronger in image-to-text models [eval]
- IC-554In Stable Diffusion 1.4, 1.5, 2.0, and 3.0, parameters with the smallest absolute values (below ~10^-3) do not contribute to the generative process, and this ineffectiveness is caused by stochastic training dynamics rather than architectural redundancy. [eval]
- IC-783Stable 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 [eval]
- IC-784Pre-trained scoring models (CLIP Score, HPS, Image Reward, Pick Score) underperform on domain-specific fine-tuned diffusion models [compared-to]
- IC-847CLIP ViT-B/32 CLIPScore achieves only ρ=0.276 / τ=0.191 correlation with human 1-5 likert T2I alignment ratings on TIFA160 [primary]
- IC-982In Stable Diffusion's UNET, visual attribute knowledge is distributed across multiple components with attribute-specific patterns, concentrated more in the up-block, and cross-attention layers are not the primary causal states [eval]
- IC-983In Stable Diffusion's CLIP text-encoder, knowledge about all visual attributes is localized to a single causal state: the first self-attention layer at the last subject token [eval]