Michael Curtis Mozer

Redirection for Erasing Memory (REM): Towards a universal unlearning method for corrupted data

Decoupling The "What" and "Where" With Polar Coordinate Positional Embedding

From Dormant to Deleted: Tamper-Resistant Unlearning Through Weight-Space Regularization

Low Compute Unlearning via Sparse Representations

On the Transfer of Object-Centric Representation Learning

Improving Discrete Optimisation Via Decoupled Straight-Through Gumbel-Softmax

AI-Assisted Generation of Difficult Math Questions

Can foundation models actively gather information in interactive environments to test hypotheses?