Patrick Schramowski

This report examines harmful prompts and safety measures in image-generating models.

The impact of language on gender stereotypes in image generation technologies.

DeiSAM enhances image understanding by combining neural networks with logical reasoning.

ALERT benchmark assesses safety risks in language models to improve their responses.

A new system assesses safety risks in images generated by AI models.

A new method improves efficiency in language processing by simplifying tokenization.

Scar enhances language models by reducing toxic language in text generation.

M-ALERT tests language models for safety across five languages.