Sotabase

Career

· Research Scientist, DeepMind2020–
· PhD in Artificial Intelligence, UC Berkeley2014–2020

Publications (31)

On the Utility of Learning about Humans for Human-AI Coordination
Neural Information Processing Systems · 2019
488
cited
Chlorophyll: synthesis-aided compiler for low-power spatial architectures
ACM-SIGPLAN Symposium on Programming Language Design and Implementation · 2014
155
cited
Does Circuit Analysis Interpretability Scale? Evidence from Multiple Choice Capabilities in Chinchilla
arXiv.org · 2023
142
cited
Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goals
arXiv.org · 2022
113
cited
Evaluating Frontier Models for Dangerous Capabilities
arXiv.org · 2024
103
cited
Optimal Policies Tend To Seek Power
Neural Information Processing Systems · 2019
96
cited
Explaining grokking through circuit efficiency
arXiv.org · 2023
75
cited
On the Feasibility of Learning, Rather than Assuming, Human Biases for Reward Inference
International Conference on Machine Learning · 2019
74
cited
Preferences Implicit in the State of the World
International Conference on Learning Representations · 2019
57
cited
Evaluating the Robustness of Collaborative Agents
Adaptive Agents and Multi-Agent Systems · 2021
41
cited
The MAGICAL Benchmark for Robust Imitation
Neural Information Processing Systems · 2020
38
cited
Benefits of Assistance over Reward Learning
2020
37
cited
Chlorophyll : Synthesis-Aided Compiler for Low-Power Spatial Architectures by Phitchaya Mangpo Phothilimthana
2015
32
cited
The MineRL BASALT Competition on Learning from Human Feedback
arXiv.org · 2021
32
cited
An Empirical Investigation of Representation Learning for Imitation
NeurIPS Datasets and Benchmarks · 2022
28
cited
SIRL: Similarity-based Implicit Representation Learning
IEEE/ACM International Conference on Human-Robot Interaction · 2023
22
cited
Active Inverse Reward Design
arXiv.org · 2018
21
cited
Choice Set Misspecification in Reward Inference
AISafety@IJCAI · 2021
19
cited
Towards Solving Fuzzy Tasks with Human Feedback: A Retrospective of the MineRL BASALT 2022 Competition
Neural Information Processing Systems · 2023
18
cited
Retrospective on the 2021 MineRL BASALT Competition on Learning from Human Feedback
Neural Information Processing Systems · 2022
10
cited
Sotabase
Rohin Shah | Researcher Profile | Sotabase | Sotabase