About Me

I am a Ph.D. candidate in Computer Science at The Hong Kong University of Science and Technology (Guangzhou), advised by Prof. Hui Xiong. I am currently working on Safe Embodied AI Pre-training within Ant Group's Security Department. Previously, I was a Research Intern and Co-Lead of core VLA research at AI² Robotics.

My research focuses on Embodied AI, particularly Vision-Language-Action models, spatial perception, understanding, and modeling. I aim to build generalizable robot intelligence that can remember, reason, and act reliably in real-world environments.

Education

HKUST (Guangzhou)
HKUST (Guangzhou)Ph.D. in Computer ScienceArtificial Intelligence Thrust · Advisor: Prof. Hui XiongSep. 2024 – Jul. 2027 (expected)
Shenzhen University
Shenzhen UniversityM.Sc. in Computer Science · GPA 3.63/4.0 · Top 1%Advisor: Prof. F. Richard YuSep. 2021 – Jun. 2024

Industry

Ant GroupSecurity DepartmentSafe Embodied AI Pre-trainingPresent
AI² RoboticsResearch Intern · Co-Lead, Core VLA ResearchCo-Founder, AlphaBrain PlatformSupervised by Prof. Yandong GuoMay 2025 – July 2026
Tencent Youtu Lab
Tencent Youtu LabComputer Vision Algorithm InternJan. 2024 – Apr. 2024
Guangming Laboratory
Guangming LaboratoryAlgorithm Engineer / Research InternAug. 2022 – Dec. 2023

Honors & Awards

  • Outstanding Graduate, Shenzhen University2024
  • National Scholarship, Ministry of Education2023
  • Reviewer for CV, robotics, and ML venuesOngoing

Research Interests

  • Vision-Language-Action Models
  • Embodied AI & Robot Learning
  • Spatial Memory & World Models
  • 3D and Event-based Perception

News

2026

SOMA is accepted to ICML 2026.

AlphaBrain reaches 180+ GitHub stars.

2025

See&Trek is accepted to NeurIPS 2025.

Joined AI² Robotics as a Research Intern.

2024

Implicit Grasp Diffusion is accepted to CoRL 2024.

Robot Trajectron is accepted to ICRA 2024.

Selected Publications (view all )

SOMA method framework

[ICML 2026] Spatial Memory for Out-of-Vision Manipulation in Vision-Language-Action

Pengteng Li, Weiyu Guo, He Zhang, Tiefu Cai, Xiao He, Yandong Guo*, Hui Xiong*

[arXiv][GitHub][Scholar]

EventVL method framework

[TMM 2026] EventVL: Understand Event Streams via Multimodal Large Language Models

Pengteng Li, Y. Lu, Pinhao Song, Wuyang Li, Huizai Yao, Hui Xiong

[arXiv][Scholar]

DeblurSplat method framework

[TMM 2026] DeblurSplat: SfM-Free 3D Gaussian Splatting with Event Camera for Robust Deblurring

Pengteng Li, Y. Lu, Pinhao Song, W. Guo, Huizai Yao, F. Richard Yu, Hui Xiong

[arXiv]

See and Trek method framework

[NeurIPS 2025] See&Trek: Training-Free Spatial Prompting for Multimodal Large Language Models

Pengteng Li, Pinhao Song, Wuyang Li, Huizai Yao, Weiyu Guo, Yijie Xu, Dugang Liu, Hui Xiong

[arXiv][GitHub]

OTOcc training framework

[IJCAI 2024] OTOcc: Optimal Transport for Occupancy Prediction

Pengteng Li, Ying He, F. Richard Yu, Pinhao Song, Xingchen Zhou, Guang Zhou

[Paper]

IGG method framework

[ACM MM 2023] IGG: Improved Graph Generation for Domain Adaptive Object Detection

Pengteng Li, Ying He, F. Richard Yu, Pinhao Song, Dongfu Yin, Guang Zhou

[Paper]

Bagging R-CNN method framework

[ICASSP 2023] Bagging R-CNN: Ensemble for Object Detection in Complex Traffic Scenes

Pengteng Li, Ying He, Dongfu Yin, F. Richard Yu, Pinhao Song

[Paper]

Boosting R-CNN method framework

[Neurocomputing 2023] Boosting R-CNN: Reweighting R-CNN Samples by RPN's Error for Underwater Object Detection

Pinhao Song, Pengteng Li, L. Dai, T. Wang, Z. Chen

[arXiv][Paper][GitHub]

All publications by category

Open Source

Open SourceAlphaBrain Platform
Co-Founder

Responsible for the core architecture design and world model design of this open-source Vision-Language-Action platform.

Website GitHub机器之心量子位
Copied