Linxi Fan

About

Linxi "Jim" Fan is a prominent robotics and AI researcher whose work sits at the intersection of robot learning, simulation, and foundation models. He has made foundational contributions to embodied AI through the development of iGibson 1.0, a richly interactive simulation environment enabling robots to learn complex household tasks across realistic large-scale scenes — work that has collectively garnered over 175 citations. His SURREAL framework (118 citations) established an influential open-source benchmark for scalable reinforcement learning in robot manipulation, while VIMA extended prompt-based learning paradigms from natural language processing into general-purpose robotic control. Fan's more recent work reflects an ambitious push toward generalizable intelligence: Eureka demonstrated that large language models can autonomously design human-level reward functions for dexterous tasks like pen spinning, and MetaMorph explored universal controllers for modular robot morphologies using Transformers. His research on humanoid whole-body control (HOVER) and bimanual dexterous manipulation (DexMimicGen) underscores his commitment to solving increasingly complex physical tasks. Across imitation learning, sim-to-real transfer, and multimodal reasoning, Fan's body of work consistently bridges cutting-edge AI with practical robotics, making him a compelling figure shaping the future of autonomous agents.

Research Focus

Key Achievements

Top Papers

1. 1
   iGibson 1.0: A Simulation Environment for Interactive Tasks in Large Realistic Scenes

   132 citations · 2021
2. 2
   SURREAL: Open-Source Reinforcement Learning Framework and Robot Manipulation Benchmark

   118 citations · 2018
3. 3
   VIMA: General Robot Manipulation with Multimodal Prompts

   65 citations · 2022
4. 4
   Eureka: Human-Level Reward Design via Coding Large Language Models

   48 citations · 2023
5. 5
   iGibson 1.0: a Simulation Environment for Interactive Tasks in Large Realistic Scenes

   45 citations · 2020
6. 6
   MimicPlay: Long-Horizon Imitation Learning by Watching Human Play

   24 citations · 2023
7. 7
   SECANT: Self-Expert Cloning for Zero-Shot Generalization of Visual Policies

   14 citations · 2021
8. 8
   HOVER: Versatile Neural Whole-Body Controller for Humanoid Robots

   12 citations · 2025
9. 9
   MetaMorph: Learning Universal Controllers with Transformers

   12 citations · 2022
10. 10
    DexMimicGen: Automated Data Generation for Bimanual Dexterous Manipulation via Imitation Learning

    8 citations · 2025

Key Collaborators

Contact & Links