Probabilistic Machine Learning Lab

Our lab develops probabilistic methods to AI model both safe and unsafe distributions, with a focus on controlling generation toward safe outcomes. We study diffusion and flow-matching frameworks for image, video, and language models, and are extending these ideas to action models. We are also deeply interested in fine-tuning foundation models, including vision-language models and large language models, while mitigating overfitting, and we consider trustworthiness and reliability as core research directions.

Research Highlights

• Probabilistic approaches for generative AI
• Safe AI grounded in probabilistic modeling
• Mitigating overfitting during post-training of foundation models