Liyiming Ke
Papers
10
Total Citations
207
H-Index
6
About
Liyiming Ke is a robotics researcher whose work sits at the intersection of robot learning, fine manipulation, and generalist robot control. Best known for contributions to the π₀ vision-language-action flow model — a landmark effort in general-purpose robot learning that has already accumulated over 127 citations since its 2025 release — Ke has consistently pushed the boundaries of what robots can physically accomplish with precision and dexterity. A distinctive thread running through Ke's research is the use of chopsticks as a testbed for fine manipulation, an elegant choice that captures the real-world complexity of grasping small, delicate, or unstable objects. This line of work, spanning teleoperation studies and imitation learning approaches that combat covariate shift, laid important groundwork for tackling broader challenges in robot dexterity. Ke has also explored reinforcement learning for delicate tasks like cherry-picking, offline reinforcement learning with realistic data, and corrective label strategies for behavior cloning. Together, these contributions reflect a researcher deeply committed to closing the gap between controlled laboratory settings and genuine real-world robot deployment — work that speaks directly to both the practical future of robotics and some of the field's deepest open questions about machine intelligence.
Research Focus
Key Achievements
Top Papers
- 1π₀: A Vision-Language-Action Flow Model for General Robot Control127 citations · 2025
- 2
- 3Cherry-Picking with Reinforcement Learning11 citations · 2023
- 4
- 5$π_0$: A Vision-Language-Action Flow Model for General Robot Control8 citations · 2024
- 6Real World Offline Reinforcement Learning with Realistic Data Source7 citations · 2023
- 7
- 8
- 9$π_{0.5}$: a Vision-Language-Action Model with Open-World Generalization2 citations · 2025
- 10