Humanoid Robot Learning
I am a second-year Ph.D. student at Fudan University and Shanghai Innovation Institute, advised by Prof. Hongyang Li and Prof. Yu Qiao. My research explores how humanoids can learn versatile loco-manipulation skills from human demonstrations and multimodal experience.
News
We introduce PingpongDuo, the first humanoid table tennis doubles system.
SMASH is accepted by IEEE Transactions on Robotics (T-RO).
EgoHumanoid is fully open-sourced and accepted by RSS 2026.
We release SMASH, mastering scalable whole-body skills for humanoid ping-pong with egocentric vision.
We release EgoHumanoid, a human-to-humanoid loco-manipulation transfer framework.
WholeBodyVLA is accepted by ICLR 2026.
Selected Projects
PingpongDuo: Agile Humanoid Table Tennis Doubles with Safe Team Coordination
The first humanoid table tennis doubles system, combining safe team coordination, agile footwork, and precise striking.
Humanoid for sports - RoboPingpong
Upgraded SMASH to a live in-the-wild humanoid ping-pong demo with stronger rallies and fully onboard computation.
SMASH: Mastering Scalable Whole-Body Skills for Humanoid Ping-Pong with Egocentric Vision
The first humanoid system to play agile ping-pong in the wild without motion capture.
EgoHumanoid: Unlocking In-the-Wild Loco-Manipulation with Robot-Free Egocentric Demonstration
The first human-to-humanoid framework for whole-body loco-manipulation from robot-free egocentric demonstrations.
WholeBodyVLA: Towards Unified Latent VLA for Whole-Body Loco-Manipulation Control
The first VLA-style humanoid system for large-space whole-body loco-manipulation.
Detect Anything 3D in the Wild
Promptable monocular 3D object detection for open-world scenes.
More Publications
Generative AI
ForgerySleuth: Empowering Multimodal Large Language Models for Image Manipulation Detection.
SketchRefiner: Text-Guided Sketch Refinement Through Latent Diffusion Models.
SAFL-Net: Semantic-Agnostic Feature Learning Network with Auxiliary Plugins for Image Manipulation Detection.
Core ContributorMachine Learning
Navigating Real-World Partial Label Learning: Unveiling Fine-Grained Images with Attributes.
First AuthorRecent Advances in Complementary Label Learning.
Corresponding AuthorComCo: Complementary Supervised Contrastive Learning for Complementary Label Learning.
First AuthorEducation
Fudan University
Ph.D. student in Computer Science
Chinese Academy of Sciences
M.S. in Machine Learning and Optimization
Hunan University
B.S. in Statistics · Rank 1/30
Honors & Service
Selected Honors
- National Scholarship
- Beijing University Basketball League · Second Place
Academic Service
Reviewer · RSS, ICLR, NeurIPS, CVPR, ICCV, AAAI, RA-L
Organizer · End-to-End Embodied Agents Workshop