Cho-Jui Hsieh

Research reveals how trigger patches influence image generation in diffusion models.

New methods strengthen watermarks against modern image editing threats.

A fresh approach to AI combines language models with symbolic programs for better interpretability.

A new benchmark tests AI models on complex math problems.

A new approach to improve text-to-image model prompts for enhanced results.

This method enhances prompt creation for large language models, improving overall performance.

New methods detect and respond to memorization in AI-generated content.

A new method to assess uncertainty in language model outputs for greater reliability.

Research shows ways to enhance context awareness in language models for better responses.