Papers
7
Total Citations
191
H-Index
5
About
Zhao Mandi is a robotics and AI researcher whose work sits at the dynamic intersection of robot learning, multi-agent systems, and large language models. Her research primarily focuses on enabling robots to generalize across diverse tasks and environments, collaborating intelligently with one another, and manipulating complex objects — including deformable materials that have long challenged the field. Mandi's most influential contribution is RoCo, a pioneering framework for multi-robot collaboration that leverages large language models to facilitate both high-level strategic dialogue and low-level path planning between robots. With 123 citations since its 2024 publication, RoCo represents a landmark step toward socially reasoning robotic systems. Her earlier work on one-shot visual imitation learning (29 combined citations) addresses the fundamental challenge of teaching robots to rapidly acquire new skills from minimal demonstrations — a critical capability for truly general-purpose machines. Her CACTI framework further advances scalable multi-task robot learning inspired by breakthroughs in vision and language AI. More recently, Mandi has explored semantically controllable data augmentation for improved generalization and scene-flow methods for deformable object manipulation via DeformGS. Across her body of work, she consistently pushes the boundaries of what robots can learn, adapt to, and accomplish collaboratively.
Research Focus
Key Achievements
Top Papers
- 1RoCo: Dialectic Multi-Robot Collaboration with Large Language Models123 citations · 2024
- 2Towards More Generalizable One-shot Visual Imitation Learning27 citations · 2022
- 3
- 4RoCo: Dialectic Multi-Robot Collaboration with Large Language Models11 citations · 2023
- 5
- 6Semantically controllable augmentations for generalizable robot learning5 citations · 2024
- 7Towards More Generalizable One-shot Visual Imitation Learning2 citations · 2021