Papers
19
Total Citations
1,007
H-Index
13
About
Joel Lehman is a leading researcher in artificial intelligence whose work has fundamentally reshaped how machines explore and learn in complex environments. His primary research areas include reinforcement learning, evolutionary computation, neuroevolution, and open-ended learning systems. Lehman is best known for co-developing the Go-Explore algorithm (2019, 228 citations) and its refinement "First Return, Then Explore" (212 citations), which revolutionized how AI agents handle hard-exploration problems—particularly in notoriously difficult Atari games like Montezuma's Revenge and Pitfall, where traditional reinforcement learning methods consistently fail. His influential work on improving exploration in evolution strategies (148 citations) demonstrated how populations of novelty-seeking agents can dramatically outperform standard approaches in deep reinforcement learning. Lehman has also made foundational contributions to diversity maintenance in deceptive domains (56 citations), evolvability search (24 citations), and neuroevolution (41 citations), including practical applications like robotic grasping with dexterous hands (21 citations). His 2023 paper "Evolution Through Large Models" (51 citations) explores the intersection of evolutionary algorithms and large-scale neural networks, continuing his trajectory of pushing the boundaries of machine intelligence.
Research Focus
Key Achievements
Top Papers
- 1Go-Explore: a New Approach for Hard-Exploration Problems228 citations · 2019
- 2First return, then explore212 citations
- 3
- 4
- 5Effective diversity maintenance in deceptive domains56 citations · 2013
- 6Evolution Through Large Models51 citations · 2023
- 7Neuroevolution41 citations · 2013
- 8Evolvability Search24 citations · 2016
- 9Grasping novel objects with a dexterous robotic hand through neuroevolution21 citations · 2014
- 10Encouraging reactivity to create robust machines20 citations · 2013