Wenli Xiao

Research

Co-Lead

Self-Improving Vision-Language-Action Models with Data Generation via Residual RL

ICLR 2026

PLD (Probe, Learn, Distill) is a plug-and-play recipe for Vision-Language-Action (VLA) post-training. It is model agnostic, supporting both autoregressive and diffusion architectures, and can push success rates to 99%.

Wenli Xiao*, Haotian Lin*, Andy Peng, Haoru Xue, Tairan He, Yuqi Xie, Fengyuan Hu, Jimmy Wu, Zhengyi Luo, Linxi "Jim" Fan†, Guanya Shi, Yuke Zhu†

Website Twitter

Co-Author

ASPIRE: Agentic Skills Discovery for Robotics

arXiv 2026

ASPIRE enables coding agents to discover reusable robot skills by inspecting execution feedback, debugging control programs, and evolving a growing skill library. Skills learned in simulation transfer to unseen tasks and real robots, reducing the cost of robot programming.

Runyu Lu*†, Yubo Wu*, Ethan Kou*, Max Fu, Wenli Xiao, Ajay Mandlekar, Yinzhen Xu, Guanya Shi, Ken Goldberg, Ang Chen, Mosharaf Chowdhury, Yuke Zhu†, Linxi "Jim" Fan†, Guanzhi Wang†

arXiv Website Code

Co-Lead1 / 16

ENPIRE: Agentic Robot Policy Self-Improvement in the Real World

CoRL 2026

Physical Autoresearch on real-world Robot Fleet. ENPIRE lets coding agents autonomously improve robot manipulation policies through a closed-loop physical feedback system—automatic environment reset and verification, parallel robot rollouts, and evolutionary refinement—reaching a 99% success rate on challenging dexterous manipulation tasks.

Wenli Xiao*, Jia Xie*, Tonghe Zhang*, Haotian Lin*, Letian "Max" Fu, Haoru Xue, Jalen Lu, Yi Yang, Cunxi Dai, Zi Wang, Jimmy Wu, Guanzhi Wang, S. Shankar Sastry, Ken Goldberg, Linxi "Jim" Fan‡, Yuke Zhu‡, Guanya Shi‡

Website Twitter Media

swipe to browse

Wenli Xiao

I am currently a Research Intern at Physical Intelligence.

I'm a final-year PhD student at CMU Robotics, advised by Prof. Guanya Shi. I am a full-stack roboticist, with experience ranging from gripper hardware design to large-scale VLA pre-training. My vision is to build scalable general intelligent robots in the real world.

I spent two wonderful years (2024-2026) interning at NVIDIA GEAR Lab, doing Foundation Model Post-training and Physical Auto-Research with Dr. Jim Fan and Prof. Yuke Zhu.

ENPIRE: Agentic Robot Policy Self-Improvement in the Real World

CoRL 2026

TL;DR: Physical Autoresearch on real-world Robot Fleet. ENPIRE lets coding agents autonomously improve robot manipulation policies through a closed-loop physical feedback system—automatic environment reset and verification, parallel robot rollouts, and evolutionary refinement—reaching a 99% success rate on challenging dexterous manipulation tasks.

ASPIRE: Agentic Skills Discovery for Robotics

arXiv 2026

TL;DR: ASPIRE enables coding agents to discover reusable robot skills by inspecting execution feedback, debugging control programs, and evolving a growing skill library. Skills learned in simulation transfer to unseen tasks and real robots, reducing the cost of robot programming.

HOVER: Versatile Neural Whole-Body Controller for Humanoid Robots

ICRA 2025

TL;DR: HOVER is a 1.5M-parameter neural network to control the body of a humanoid robot. It takes a lot of subconscious processing for us humans to walk, maintain balance, and maneuver our arms and legs into desired positions. We capture this 'subconsciousness' in HOVER, a single model that learns how to coordinate the motors of a humanoid robot to support locomotion and manipulation.

CaP-X: Benchmarking and Improving Coding Agents for Robot Manipulation

ICML 2026

TL;DR: CaP-X introduces CaP-Agent0, a training-free agentic framework enabling off-the-shelf LLMs to perform robotic manipulation via code generation, and CaP-Bench, a comprehensive evaluation suite of 100+ tasks. Frontier models achieve 30%+ zero-shot success on unseen tasks with 18% on perturbed tasks (vs. 0% for VLA models), and a 7B model improves from 20% to 72% in sim with 84% real-world transfer.

Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion

RSS 2024 (Outstanding Student Paper Award Finalist - Top 3)

ICRA 2024, Agile Robotics Workshop (Spotlight)

TL;DR: Legged robots navigating cluttered environments must be jointly agile for efficient task execution and safe to avoid collisions with obstacles or humans. Existing studies either develop conservative controllers (< 1.0 m/s) to ensure safety, or focus on agility without considering potentially fatal collisions. This paper introduces Agile But Safe (ABS), a learning-based control framework that enables agile and collision-free locomotion for quadrupedal robots.

Safe Deep Policy Adaptation

ICRA 2024

CoRL 2023 Deployable Workshop

TL;DR: This paper jointly tackles policy adaptation and safe reinforcement learning with safety guarantees. Comprehensive experiments on (1) classic control problems (Inverted Pendulum), (2) simulation benchmarks (Safety Gym), and (3) a real-world agile robotics platform (RC Car) demonstrate great superiority of SafeDPA in both safety and task performance, over state-of-the-art baselines.

Research

Self-Improving Vision-Language-Action Models with Data Generation via Residual RL

ASPIRE: Agentic Skills Discovery for Robotics

ENPIRE: Agentic Robot Policy Self-Improvement in the Real World

News

Research Projects

Authors

ENPIRE: Agentic Robot Policy Self-Improvement in the Real World

Authors

ASPIRE: Agentic Skills Discovery for Robotics

Authors

Self-Improving Vision-Language-Action Models with Data Generation via Residual RL

Authors

ASAP: Aligning Simulation and Real-World Physics

Authors

HOVER: Versatile Neural Whole-Body Controller for Humanoid Robots

Authors

AnyCar to Anywhere: Learning Universal Dynamics Model

Authors

CaP-X: Benchmarking and Improving Coding Agents for Robot Manipulation

Authors

DoorMan: Opening the Sim-to-Real Door for Humanoid Pixel-to-Action Policy Transfer

Authors

VIRAL: Visual Sim-to-Real at Scale for Humanoid Loco-Manipulation

Authors

SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control

Authors

Hold My Beer: Learning Gentle Humanoid Locomotion and End-Effector Stabilization Control

Authors

OmniH2O: Universal and Dexterous Human-to-Humanoid Whole-Body Teleoperation and Learning

Authors

WoCoCo: Learning Whole-Body Humanoid Control with Sequential Contacts

Authors

Learning Human-to-Humanoid Real-Time Whole-Body Teleoperation

Authors

Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion

Authors

Safe Deep Policy Adaptation

Background