Ada is a PhD student in the group, initally workin on global explanations for neural models such as feature visualizations. More recently, she has been investigating the representations learned by image-generating diffusion models, understanding when, where, and how concepts emerge to improve down-stream tasks and solve common failure modes, such as complex compositional generation.