Pengxiang Ding

Researcher & Founder

Pengxiang Ding 丁鹏翔

I am the founder and CEO of Symbiosis Robotics (共生知行), where we build foundation models for whole-body humanoid intelligence.

My research focuses on vision-language-action models and whole-body control.

I am a Ph.D. student at Zhejiang University in a joint program with Westlake University, advised by Prof. Donglin Wang in the Machine Intelligence Laboratory (MiLAB).

Previously, I received my M.Sc. from the School of Artificial Intelligence at BUPT in 2022, advised by Prof. Jianqin Yin.

Email Google Scholar GitHub X ORCID
Pengxiang Ding as a child
Hangzhou, China
Symbiosis Robotics共生知行
From Humanoid to Human
Google Scholar 2,312 Total citations As of

News

Latest updates
Scroll for more

Publications

Vision, language
& robot learning

Google Scholar

Selected publications 21

2026

First author or project leader. Click a figure to enlarge.
* Equal contribution   † Project leader

NeurIPS

Endowing Your Vision-Language-Action Model with a Predictive Mind

Pengxiang Ding, Haoying Wang, Minghui Lin, Qishen Wang, Zhenyu Ding, Runze Suo, Xuanxuan An, Wenxuan Song, Fuhao Li, Han Zhao, Donglin Wang, Ning Ding

Paper Conference
CoRL

SAGE-3D: Semantic-Aware 3D Representations for Generalizable Vision-Language-Action Models

Pengxiang Ding, Xuanxuan An, Kexian Yu, Haoying Wang, Minghui Lin, Xuhao Wan, Wenxuan Song, Han Zhao, Yihao Wang, Donglin Wang, Ning Ding

ECCV

Towards More Efficient Decoding for Autoregressive Vision-language-action Models

Consistency distillation and early-exit decoding accelerate autoregressive VLA inference.

Wenxuan Song*, Jiayi Chen*, Pengxiang Ding†, Yuxin Huang, Han Zhao, Yinchuan Li, Yingcong Chen, Donglin Wang, Haoang Li

Paper Conference Project
CEED-VLA framework showing Jacobi trajectory collection, consistency distillation, and early-exit decoding. Method overview
CVPR

HiF-VLA: Hindsight, Insight and Foresight through Motion Representation for Vision-Language-Action Models

Motion-based hindsight and foresight support temporally coherent, long-horizon manipulation.

Minghui Lin, Pengxiang Ding†, Shu Wang, Zifeng Zhuang, Yang Liu, Xinyang Tong, Wenxuan Song, Shangke Lyu, Siteng Huang†, Donglin Wang

Paper Project Code
HiF-VLA robot gripper placing a white block on a plate on a green tabletop. Demo screenshot
CVPR

CUBic: Coordinated Unified Bimanual Perception and Control Framework

A shared perceptual representation connects independent arm observations with coordinated bimanual control.

Xingyu Wang*, Pengxiang Ding*†, Jingkai Xu, Donglin Wang, Zhaoxin Fan

Paper arXiv
CUBic framework connecting arm-specific perception and shared codebooks to a bimanual diffusion policy. Method overview
ICLR

Unified Diffusion VLA: Vision-Language-Action Model via Joint Discrete Denoising Diffusion Process

Future visual observations and robot actions are refined together through a joint discrete diffusion process.

Jiayi Chen*, Wenxuan Song*, Pengxiang Ding†, Ziyang Zhou, Han Zhao, Feilong Tang, Donglin Wang, Haoang Li

Paper Project Code
UD-VLA architecture showing unified image, text, and action tokens and joint discrete denoising. Method overview
ICRA

TrajBooster: Boosting Humanoid Whole-Body Manipulation via Trajectory-Centric Learning

Retargeted end-effector trajectories transfer wheeled-humanoid data to bipedal whole-body manipulation.

Jiacheng Liu*, Pengxiang Ding†*, Qihang Zhou, Yuxuan Wu, Da Huang, Zimian Peng, Wei Xiao, Weinan Zhang, Lixin Yang, Cewu Lu, Donglin Wang

Paper Project Code
Unitree G1 humanoid manipulating a toy beside a storage basket in a TrajBooster demonstration. Demo screenshot
AAAI

VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Model

Bridge Attention connects a compact vision-language backbone to an action policy.

Yihao Wang*, Pengxiang Ding†*, Lingxiao Li, Can Cui, Zirui Ge, Xinyang Tong, Wenxuan Song, Han Zhao, Wei Zhao, Pengxu Hou, Siteng Huang, Yifan Tang, Wenhui Wang, Ru Zhang, Jianyi Liu, Donglin Wang

Paper Project Code
VLA-Adapter architecture with Bridge Attention linking the vision-language model and action policy. Method overview
Earlier selected work13 papers · 2022–2025
CoRL

Long-VLA: Unleashing Long-Horizon Capability of Vision Language Action Model for Robot Manipulation

Yiguo Fan†, Pengxiang Ding†,Shuanghao Bai†, Xinyang Tong†, Yuyang Zhu, Hongchao Lu, Fengqi Dai, Wei Zhao, Yang Liu, Siteng Huang, Zhaoxin Fan, Badong Chen, Donglin Wang

ICCV

CARP: Visuomotor Policy Learning via Coarse-to-Fine Autoregressive Prediction

A coarse-to-fine autoregressive policy progressively refines multi-scale action sequences.

Zhefei Gong*, Pengxiang Ding†, Shangke Lyu, Siteng Huang, Mingyang Sun, Wei Zhao, Zhaoxin Fan, Donglin Wang

Paper Webpage Code Article (中文)
Comparison of autoregressive and diffusion policies with CARP coarse-to-fine action generation. Method overview
NeurIPS

SSR: Enhancing Depth Perception in Vision-Language Models via Rationale-Guided Spatial Reasoning

Yang Liu†, Ming Ma†, Xiaomin Yu†, Pengxiang Ding†, Han Zhao, Mingyang Sun, Siteng Huang, Donglin Wang

Paper Webpage Code
arXiv

Humanoid-vla: Towards universal humanoid control with visual integration

Pengxiang Ding†, Jianfei Ma†, Xinyang Tong†, Binghong Zou, Xinxin Luo, Yiguo Fan, Ting Wang, Hongchao Lu, Panzhong Mo, Jinxin Liu, Yuefan Wang, Huaicheng Zhou, Wenshuo Feng, Jiacheng Liu, Siteng Huang, Donglin Wang

Paper
IROS

Accelerating vision-language-action model integrated with action chunking via parallel decoding

Wenxuan Song†, Jiayi Chen†, Pengxiang Ding†, Han Zhao, Wei Zhao, Zhide Zhong, Zongyuan Ge, Jun Ma, Haoang Li

Paper
ICML

Rethinking Latent Representations in Behavior Cloning:An Information Bottleneck Approach for Robot Manipulation

Shuanghao Bai*, Wanqi Zhou*, Pengxiang Ding†, Wei Zhao, Donglin Wang, Badong Chen

Paper Webpage Code
ICRA

QUART-Online: Latency-Free Large Multimodal Language Model for Quadruped Robot Learning

Action chunk discretization enables faster multimodal policies for quadruped robots.

Xinyang Tong†, Pengxiang Ding†, Donglin Wang, Wenjie Zhang, Can Cui, Mingyang Sun, Yiguo Fan, Han Zhao, Hongyin Zhang, Yonghao Dang, Siteng Huang, Shangke Lyu

Paper Webpage
QUART-Online teaser showing compressed action tokens and a quadruped navigating an obstacle. Method overview
ICLR

GEVRM: Goal-Expressive Video Generation Model For Robust Visual Manipulation

Hongyin Zhang, Pengxiang Ding†, Shangke Lyu, Ying Peng, Donglin Wang

Paper
ICLR

VLAS: Vision-Language-Action Model with Speech Instructions for Customized Robot Manipulation

Wei Zhao, Pengxiang Ding†, Zhang Min, Zhefei Gong, Shuanghao Bai, Han Zhao, Donglin Wang

Paper
ECCV

QUAR-VLA: Vision-Language-Action Model for Quadruped Robots

Visual observations and language instructions guide quadruped perception, navigation, and action.

Pengxiang Ding, Han Zhao, Wenxuan Song, Wenjie Zhang, Min Zhang, Siteng Huang, Ningxi Yang, Donglin Wang

Paper Project
QUAR-VLA overview from simulated and real robot data to vision-language-action training and deployment. Method overview
ECCV

PiTe: Pixel-Temporal Alignment for Large Video-Language Model

Yang Liu†, Pengxiang Ding†, Siteng Huang, Min Zhang, Han Zhao, Donglin Wang

Paper Code
AAAI

Expressive Forecasting of 3D Whole-body Human Motions

Pengxiang Ding, Qiongjie Cui, Min Zhang, Mengyuan Liu, Haofan Wang, Donglin Wang

Paper Code
TCSVT

Towards more realistic human motion prediction with attention to motion coordination

Pengxiang Ding, Jianqin Yin

Paper Code
Other publications8 papers · 2026
ECCV

Fast-dVLA: Accelerating Discrete Diffusion VLA to Real-Time Performance

Wenxuan Song, Jiayi Chen, Shuai Chen, Jingbo Wang, Pengxiang Ding, Han Zhao, Yikai Qin, Xinhu Zheng, Yan Wang, Donglin Wang, Haoang Li

Paper Conference
ICML

Dyn-VPP: Video Prediction Policy Optimization for Improved Visual Dynamics

Zirui Ge, Pengxiang Ding, Baohua Yin, Yemin Wang, Qishen Wang, Zhiyong Xie, Hengtao Li, Runze Suo, Wenxuan Song, Han Zhao, Shangke Lyu, Haoang Li, Ran Cheng, Cheng Chi, Hui-Bin Ge, Yaozhi Luo, Donglin Wang

Paper Openreview
ICRA

Rethinking the Practicality of Vision-Language-Action Model: A Comprehensive Benchmark and an Improved Baseline

Wenxuan Song, Jiayi Chen, Xiaoquan Sun, Huashuo Lei, Yikai Qin, Wei Zhao, Pengxiang Ding, Han Zhao, Tongxin Wang, Pengxu Hou, Zhide Zhong, Haodong Yan, Donglin Wang, Jun Ma, Haoang Li

Paper
ICRA

Robust Online Residual Refinement via Koopman-Guided Dynamics Modeling

Zhefei Gong, Shangke Lyu, Pengxiang Ding, Wei Xiao, Donglin Wang

Paper
ICLR

Spatial Forcing: Implicit Spatial Representation Alignment for Vision-language-action Model

Fuhao Li, Wenxuan Song, Han Zhao, Jingbo Wang, Pengxiang Ding, Donglin Wang, Long Zeng, Haoang Li

Paper Webpage Code
AAAI

ReconVLA: Reconstructive Vision-Language-Action Model as Effective Robot Perceiver

Wenxuan Song, Ziyang Zhou, Han Zhao, Jiayi Chen, Pengxiang Ding, Haodong Yan, Yuxin Huang, Feilong Tang, Donglin Wang, Haoang Li

Paper Interview (中文)
AAAI

Filter, Correlate, Compress: Training-Free Token Reduction for MLLM Acceleration

Yuhang Han, Xuyang Liu, Zihan Zhang, Pengxiang Ding, Junjie Chen, Honggang Chen, Donglin Wang, Qingsen Yan, Siteng Huang

Paper
AAAI

Rethinking Target Label Conditioning in Adversarial Attacks: A 2D Tensor-Guided Generative Approach

Hangyu Liu, Bo Peng, Pengxiang Ding, Donglin Wang

Paper

Experience

Research & industry

  1. –

    Xiaomi 小米

    Intern

  2. –

    Ant Group · Robbyant 蚂蚁灵波

    Intern

  3. Jan 2025–Jun 2025

    DAMO Academy

    Research Intern · Machine Intelligence Laboratory

    Advisor: Xin Li

  4. Sep 2022–Mar 2023

    RedBook

    Research Intern · Intelligent Creation Group

    Advisor: Haofan Wang

  5. Sep 2021–Mar 2022

    SenseTime

    Research Intern · Smart City Group

    Advisor: Dongliang Wang

Talks

Invited talks
& discussions

  1. Talk

    What Are AI Builders Changing? 

    Chongli Forum · AI Builders Show · 甲子光年

  2. Panel discussion

    World Models, Robot Intelligence, and Control 

    WRC 2026 Developer After Party · 北京人形 / Xbotics

  3. Conference talk

    Advancing VLA Models across Robot Embodiments 

    EAIRCon 2025 · Shenzhen · 智猩猩 / 智东西

  4. Joint session

    VLA Foundations, Reinforcement Learning, and Dexterous Manipulation 

    Lumina Talk #19 · Westlake MiLAB

  5. Online talk

    Full-Stack VLA: From Visuomotor Policies to Robot Foundation Models 

    3D Vision Workshop · 3D 视觉工坊

  6. Open lecture

    End-to-End Foundation Models for Quadruped Robots 

    Shenlan Academy · 深蓝学院

  7. Research talk

    Vision-Language-Action Models

    Peking University · Hosted by Hao Dong

In the media

Interviews & features

Service

Academic community

Journal & conference reviewer

ICML, ICLR, NeurIPS, CVPR, ICCV, ACM MM, AAAI, ICRA, IROS, CoRL, TNNLS, TASE, TCSVT

You can also find me on Redbook .

View source