Mara Levy
Papers
2
Total Citations
5
H-Index
2
About
Mara Levy is a robotics researcher advancing the frontier of data-efficient robot learning, with a focus on enabling machines to generalize across diverse environments with minimal human input. Her work centers on two key challenges: learning from sparse demonstrations and achieving robust visuo-spatial generalization. In her 2025 paper "P3-PO: Prescriptive Point Priors for Visuo-Spatial Generalization of Robot Policies" (3 citations), she tackles the persistent problem of policy brittleness by introducing structured priors that allow robots to adapt to varied object instances and environmental conditions without massive datasets. Her earlier work, "WayEx: Waypoint Exploration using a Single Demonstration" (2024, 2 citations), proposes a novel method for learning complex goal-conditioned tasks from just one expert example—dramatically reducing the data burden compared to traditional imitation learning. By eliminating the need for action information during demonstrations, WayEx opens new possibilities for non-experts to teach robots. Though early in her career, Levy’s focus on sample efficiency and generalization positions her at the forefront of making robot learning more accessible and practical for real-world deployment.
Research Focus
Key Achievements
Top Papers
- 1
- 2WayEx: Waypoint Exploration using a Single Demonstration2 citations · 2024