Profiles2020-2025
Anca D Dragan
University of California, Berkeley
Also published as Anca Dragan
16Papers
8Last author
25Unique co-authors
0First author
0Solo
First and last author come from byline position, which some venues assign alphabetically.
Recent papers16
- Planning without Search: Refining Frontier LLMs with Offline Goal-Conditioned RLAuthor 2 of 3NeurIPS 2025 · 11 citations
- Adversaries Can Misuse Combinations of Safe ModelsAuthor 2 of 4ICML 2025
- Correlated Proxies: A New Definition and Improved Mitigation for Reward HackingAuthor 3 of 3ICLR 2025
- Q-SFT: Q-Learning for Language Models via Supervised Fine-TuningAuthor 2 of 3ICLR 2025
- Offline RL with Observation Histories: Analyzing and Improving Sample ComplexityAuthor 2 of 4ICLR 2024 · 8 citations
- Automatically Auditing Large Language Models via Discrete OptimizationAuthor 2 of 4ICML 2023 · 232 citations
- The Effect of Modeling Human Rationality Level on Learning Rewards from Multiple Feedback TypesAuthor 4 of 4AAAI 2023 · 38 citations
- Learning to Influence Human Behavior with Offline Reinforcement LearningAuthor 3 of 4NeurIPS 2023 · 30 citations
- Inferring Rewards from Language in ContextAuthor 4 of 4ACL 2022 · 71 citations
- The Boltzmann Policy Distribution: Accounting for Systematic Suboptimality in Human ModelsAuthor 2 of 2ICLR 2022 · 46 citations
- First Contact: Unsupervised Human-Machine Co-Adaptation via Mutual Information MaximizationAuthor 3 of 3NeurIPS 2022 · 18 citations
- Pragmatic Image Compression for Human-in-the-Loop Decision-MakingAuthor 2 of 3NeurIPS 2021 · 16 citations
- Learning What To Do by Simulating the PastAuthor 4 of 4ICLR 2021 · 4 citations
- SQIL: Imitation Learning via Reinforcement Learning with Sparse RewardsAuthor 2 of 3ICLR 2020 · 299 citations
- Reward-rational (implicit) choice: A unifying formalism for reward learningAuthor 3 of 3NeurIPS 2020 · 219 citations
- AvE: Assistance via EmpowermentAuthor 6 of 6NeurIPS 2020 · 51 citations
