Research

A survey on data attribution with focus on generative AI.

We design the first efficient machine unlearning evaluation metric with provable guarantees.

We propose an efficient gradient compression algorithm to accelerate and scale gradient-based data attribution methods to billion-scale models.

We consider the adversarial attack on training data attribution methods.

We developed a efficient library for data attribution, aiming to streamline the development of data attribution algorithms.

We provide a comprehensive study of the common practices in the Most Influential Subset Selection (MISS) problem.

We propose a new non-linear data augmentation framework powered by information geometry.

Exploring a novel approach to exactly solve an NP-hard combinatorial optimization problem by using imitation learning.