 A picture of the space of typical learnable tasksInternational Conference on Machine Learning (ICML), 2022 |
 Towards Understanding Grokking: An Effective Theory of Representation
LearningNeural Information Processing Systems (NeurIPS), 2022 |
 Deep Reference Priors: What is the best way to pretrain a model?International Conference on Machine Learning (ICML), 2022 |
 Controllable Guarantees for Fair Outcomes via Contrastive Information
EstimationAAAI Conference on Artificial Intelligence (AAAI), 2021 |
 An Information-Geometric Distance on the Space of TasksInternational Conference on Machine Learning (ICML), 2020 |
