Sotabase

Career

· CEO & Co-Founder, FAR.AI2022–

Publications (42)

Adversarial Policies: Attacking Deep Reinforcement Learning
International Conference on Learning Representations · 2019
405
cited
This Paper Is Included in the Proceedings of the 12th Usenix Symposium on Operating Systems Design and Implementation (osdi '16). Firmament: Fast, Centralized Cluster Scheduling at Scale Firmament: Fast, Centralized Cluster Scheduling at Scale
159
cited
Multi-Agent Risks from Advanced AI
arXiv.org · 2025
88
cited
Quantifying Differences in Reward Functions
International Conference on Learning Representations · 2020
73
cited
Invariance in Policy Optimisation and Partial Identifiability in Reward Learning
International Conference on Machine Learning · 2022
53
cited
Inverse reinforcement learning for video games
arXiv.org · 2018
50
cited
imitation: Clean Imitation Learning Implementations
arXiv.org · 2022
46
cited
Multi-task Maximum Entropy Inverse Reinforcement Learning
arXiv.org · 2018
41
cited
Understanding Learned Reward Functions
arXiv.org · 2020
40
cited
Uncertainty Estimation for Language Reward Models
arXiv.org · 2022
36
cited
Adversarial Policies Beat Superhuman Go AIs
International Conference on Machine Learning · 2022
33
cited
Adversarial Policies Beat Professional-Level Go AIs
arXiv.org · 2022
28
cited
Exploiting Novel GPT-4 APIs
arXiv.org · 2023
28
cited
Scaling Trends for Data Poisoning in LLMs
AAAI Conference on Artificial Intelligence · 2024
24
cited
On The Fragility of Learned Reward Functions
arXiv.org · 2023
23
cited
Active Inverse Reward Design
arXiv.org · 2018
21
cited
A Primer on Maximum Causal Entropy Inverse Reinforcement Learning
arXiv.org · 2022
18
cited
STARC: A General Framework For Quantifying Differences Between Reward Functions
International Conference on Learning Representations · 2023
12
cited
Scaling Trends in Language Model Robustness
2024
11
cited
The Singapore Consensus on Global AI Safety Research Priorities
Robotics · 2025
10
cited
Sotabase
Adam Gleave | Researcher Profile | Sotabase | Sotabase