CV

PhD, Joint Doctoral Program

Shanghai Jiao Tong University & Eastern Institute of Technology, Ningbo

Computer Science and Technology

• Human-centered embodied intelligence, humanoid self-representation, reliable embodied-agent systems, 3D vision, and humanoid motion and manipulation.

Master of Science

Tsinghua University

Data Science and Information Technology

• Research in 3D vision, geometry processing, and machine learning.

Bachelor of Engineering

Dalian University of Technology

Software Engineering

Humanoid Bodily Self-Representation

Learning self-other distinction and predictive 3D body models directly from proprioceptive-visual experience.

• Developed a self-supervised framework that identifies the robot body without identity labels or a predefined kinematic model.
• Bootstrapped a 3D occupancy self-model for target reaching, collision-aware motion planning, and human-to-robot motion retargeting.

Thea Embodied-Agent Harness & Unitree G1

Closed-loop tool use for long-horizon robot tasks in real environments.

• Led the deployment, integration, and debugging of Unitree G1 locomotion, grasping, and pick-and-place capabilities.
• Contributed to a harness that unifies language models, robot policies, and physical tools with dynamic task composition and failure recovery.
• Contributed to persistent scene-graph state and independent evaluators that classify actions as successful, failed, or in progress.

Interpretable Scientific Machine Learning with KAN

Physics-guided learning of explicit formulas for large-scale soil-moisture retrieval.

• Combined a radiative transfer model with a Kolmogorov-Arnold Network to replace black-box prediction with inspectable symbolic relationships.
• Achieved a Pearson correlation of 0.98 on non-forest in-situ data while improving interpretability and generalization over an MLP.

Generative 3D Rigging and Shape Correspondence

Learning controllable structures for generated 3D assets.

• Co-authored ARMO, DRiVE, and SRIF on autoregressive skeleton generation, diffusion-based character rigging, and semantic shape registration.

Proprioceptive-visual Correspondence Enables Self-other Distinction in Humanoid Robots

arXiv preprint

A Novel Soil Moisture Retrieval Method via Combining Radiative Transfer Model and Machine Learning

Remote Sensing of Environment · JIF 12.3 (2025) · JCR Q1 (SCIE) · CAS Q1 (Top)

Towards the Harness of Embodied Agents

arXiv preprint · Technical Report

ARMO: Autoregressive Rigging for Multi-Category Objects

ICCV

DRiVE: Diffusion-based Rigging Empowers Generation of Versatile and Expressive Characters

CVPR

SRIF: Semantic Shape Registration Empowered by Diffusion-based Image Morphing and Flow Estimation

SIGGRAPH Asia

Is Nighttime Light Primarily From Human Settlements? Exploring the Spatial Relationship Between NTL and Impervious Surface in Zhejiang Province, China

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · CAS Q2

The Passive Microwave Remote Sensing in Soil Moisture Retrieval: Products, Models, Applications and Challenges

International Soil and Water Conservation Research · CAS Q1 (Top)

National Scholarship (twice)

Ministry of Education of China

Awarded within the top 4%.

CSSC Scholarship

Dalian University of Technology

University-level high-value scholarship, top 0.5%.

First Prize, Asia and Pacific Mathematical Contest in Modeling

Top 5%.

Outstanding Graduate of Liaoning Province