Jacky Liang

About

Jacky Liang is a robotics researcher whose work spans robot manipulation, grasp planning, and the intersection of large language models (LLMs) with embodied AI. He is perhaps best known for his foundational contributions to the Dex-Net project, particularly Dex-Net 2.0, which demonstrated that deep learning models trained on massive synthetic datasets of point clouds and analytic grasp metrics could enable robots to plan robust grasps reliably — a paper that has accumulated over 1,100 citations and helped define modern data-driven robot grasping. His research on gripper tip design and GPU-accelerated simulation further reflects his commitment to practical, scalable robotic systems. More recently, Liang has made significant strides in applying LLMs to robot control. His work on "Code as Policies" (561 citations) showed that code-generating language models could be repurposed to write executable robot policy code from natural language commands, opening a compelling new paradigm for flexible robot programming. His contributions to "Inner Monologue" and the large-scale Open X-Embodiment initiative further establish him as a key figure in generalizable robot learning. Across more than a decade of research, Liang's work consistently bridges the gap between learning-based methods and real-world robotic deployment, earning him broad recognition within the robotics and AI communities.

Research Focus

Key Achievements

Top Papers

1. 1
   Dex-Net 2.0: Deep Learning to Plan Robust Grasps with Synthetic Point Clouds and Analytic Grasp Metrics

   1,162 citations · 2017
2. 2
   Code as Policies: Language Model Programs for Embodied Control

   561 citations · 2023
3. 3
   Dex-Net 2.0: Deep Learning to Plan Robust Grasps with Synthetic Point Clouds and Analytic Grasp Metrics

   277 citations · 2017
4. 4
   Inner Monologue: Embodied Reasoning through Planning with Language Models

   206 citations · 2022
5. 5
   DexPilot: Vision-Based Teleoperation of Dexterous Robotic Hand-Arm System

   197 citations · 2020
6. 6
   Open X-Embodiment: Robotic Learning Datasets and RT-X Models : Open X-Embodiment Collaboration⁰

   119 citations · 2024
7. 7
   Open X-Embodiment: Robotic Learning Datasets and RT-X Models

   101 citations · 2023
8. 8
   GPU-Accelerated Robotic Simulation for Distributed Reinforcement\n Learning

   74 citations · 2018
9. 9
   Design of parallel-jaw gripper tip surfaces for robust grasping

   50 citations · 2017
10. 10
    Code as Policies: Language Model Programs for Embodied Control

    36 citations · 2022

Key Collaborators

Contact & Links