David Klee
Papers
3
Total Citations
11
H-Index
2
About
David Klee is a rising researcher at the forefront of robotic manipulation, with a focus on sample-efficient learning through the integration of equivariance and structured world models. His work addresses a core challenge in robotics: enabling agents to learn complex manipulation tasks with minimal real-world interaction. Klee’s most cited paper, “SEIL: Simulation-augmented Equivariant Imitation Learning” (2023, 6 citations), pioneers the use of symmetry-preserving data augmentation to dramatically improve sample efficiency in imitation learning, bridging the simulation-to-reality gap. In “Factored World Models for Zero-Shot Generalization in Robotic Manipulation” (2022, 3 citations), he tackles the combinatorial explosion of object interactions by decomposing world models into object-level factors, allowing for robust zero-shot generalization in pick-and-place tasks. His latest work, “Equivariant Reinforcement Learning under Partial Observability” (2024, 2 citations), extends these principles to partially observable domains, demonstrating how symmetry-based inductive biases can unlock efficient learning where traditional methods struggle. With a total of 11 citations across his early-career publications, Klee is establishing himself as a key innovator in building more intelligent, data-efficient robotic systems.
Research Focus
Key Achievements
Top Papers
- 1SEIL: Simulation-augmented Equivariant Imitation Learning6 citations · 2023
- 2
- 3Equivariant Reinforcement Learning under Partial Observability2 citations · 2024