Modelpedia
Work in progress
About
Findings
Models
Concepts
Methods
Datasets
Sources
Light
Dark
Precise Parameter Localization for Textual Generation in Diffusion Models
2025-01-22
· ICLR 2025 Poster ·
anchor
Findings
IC-394
Text generation in SDXL, DeepFloyd IF, and SD3 is controlled by less than 1% of parameters concentrated in specific cross- or joint-attention layers, and these layers are specialised for text content rather than visual template