diff --git a/index.html b/index.html index 1c8697a..ee987d6 100644 --- a/index.html +++ b/index.html @@ -249,7 +249,7 @@

Compositional Robustness

Multilinguality

method

Evaluating the alignment of prompt embeddings as well as generated images across multiple languagese we show that well aligned embeddings enable the transfer of multiligualism to downstream tasks even for task-specific monolingual training data.

- + method
@@ -258,6 +258,7 @@

Attention Manipulation for Multimodal inference

Attention Manipulation allows us to weight image and text tokens at inference time and guide their influence on the resulting generation.

+ method
@@ -267,14 +268,19 @@

Applications

Interleaved multilingual, multimodal prompting

+ method

Image Composition

MultiFusion increases expresivness in composition through arbitrary and flexible promptin of image and text sequences.

+ method

Negative Prompting

Negative prompting with images enables a more powerful supression than through text prompts.

+ method

Style Modification

MultiFusion enables simple style transfer through one reference image capturing all the facets of a unique style such as color pallette, composition contrast, etc. making elaborate prompts obsolete. Additionally, MultiFusion enables highly individual prompting such as "in the style of a picture I drew". + method

Image Variation

MultiFusion produces meaningful image variations without the need for inversion or renoising if the input image.

+ method