Shutong Liu

Hi there! I'm an undergraduate student at Beihang University, majoring in Computer Science and Technology.

My research interests lie in embodied AI and foundation model pretraining, particularly humanoid foundation models, world-action models, and reinforcement learning for whole-body control. I also work on trustworthy agents and multimodal long-context reasoning.

Shutong Liu

News

Selected Publications

My research spans embodied intelligence, trustworthy agents, and security. Selected work is highlighted.

Main method overview of RLDexAgents showing hand manifolds, skill distillation, agent tooling, and pose-skill expansion.

An Extensible Dexterous Skill Toolkit for LLM Agents through Manifold-Guided Skill Grounding

Ruoqu Chen1, Shutong Liu2, Feixiang Ruan3, Weili Fu1, Zihao Wang1, Mengdi Xu4†

1 Tsinghua University · 2 Beihang University · 3 Tongji University · 4 Tsinghua University · † Corresponding author

Technical report · October 2026

A modular toolkit that organizes human and robot hand motions into structured pose and skill banks, grounds them with reinforcement learning, and distills executable closed-loop controllers that LLM agents can retrieve, compose, and extend.

Uni-VLaT overview with whole-body tactile sensing, multimodal prediction, and five humanoid tasks.

Uni-VLaT: Whole-Body Tactile Adaptation of VLA Policies for Humanoid Loco-Manipulation

Zihao Wang*, Shutong Liu*, Siqi Zheng, Liu Cao, Ruoqu Chen, Rundong Liu, Yanchao Yang, Mengdi Xu†

* Equal contribution · † Corresponding author

Under review at IEEE ICRA 2027

Accepted to the CoRL 2026 Workshop on Contact-Rich Loco-Manipulation

Adapting pretrained VLA policies with whole-body tactile sensing and tactile-anchored multimodal prediction for contact-rich humanoid control. Achieves 75% average success across five real-robot tasks.

S-BFP system overview showing randomized challenges and device identity calculation.

From Frozen Pose to Live Dance: Stochastic Browser Fingerprinting for Robust Risk-Based Authentication

Shutong Liu · First author

Under review at USENIX Security 2027

Stochastic browser fingerprinting introduces randomized rendering challenges to defend risk-based authentication against replay attacks while preserving stable device identification.

e-PoG overview showing the agent, adaptive monitor, fixed verifier, and corresponding extensive-form game tree.

Mitigating Agentic Deception via Evaluation-Side Partially Observable Games

Hanxiao Li1,2, Shutong Liu1, Yongjia Ma3, Jiaming Ji6, Juntao Dai6, Shuning Zhang4, Kejian Shi5, Kai Chen5, Pu Feng2, Yuqing Ma1, Qi Dou5, Yaodong Yang6, Xianglong Liu1,2,7, Simin Li1,5†

1 SKLSDE Lab, Beihang University · 2 Zhongguancun Laboratory · 3 Li Auto Inc. · 4 Tsinghua University · 5 The Chinese University of Hong Kong · 6 Institute of Artificial Intelligence, Peking University & BigAI · 7 Institute of Data Space, Hefei Comprehensive National Science Center · † Corresponding author

Under review at ICLR 2027

Formulates deception mitigation as an evaluation-side partially observable game (e-PoG) between a learning agent, an adaptive monitor, and a fixed verifier. Two-timescale policy optimization approximates a Subgame-Perfect Oversight Equilibrium, reducing measured deception by 26.0 percentage points relative to the strongest baselines while raising task success by up to 39.8 points, including under weak-to-strong supervision.

HIPPOGIST architecture with visual source traces, semantic gists, and outcome-supervised training.

HIPPOGIST: Dual-Store Memory for Long-Horizon LLM Agents with Visual Traces and Semantic Gists

Shutong Liu

Beihang University

A dual-store architecture combining compressed visual source traces and semantic gists for chunk-by-chunk long-context reasoning, with outcome-supervised memory updates. Experiments examine textual-memory and optical-memory workflows on LongBench v2.

Safety and security evolution timeline of satellite communications.

Safety and Security of Satellite Communications: A Comprehensive Review of Radio-Frequency and Free-Space Optical Systems

Weizhe Liu, Shujiang Wu, Shutong Liu, Qiang Wei, Yunhong Wang

Submitted to Security and Safety

A review of safety and security in radio-frequency and free-space optical satellite communications, organizing threats, countermeasures, and open research directions.

Education

Beihang University emblem

Beihang University

B.Eng. in Computer Science and Technology

GPA: 3.87/4.0 · Rank: 3/198 (Top 1.5%)

Research Experience

Tsinghua University emblem

Tsinghua University

Research collaborator

Whole-body tactile sensing and humanoid VLA policies for contact-rich loco-manipulation, including tactile representation learning, real-robot data collection, policy training, and deployment on Unitree G1.

Beihang University emblem

Beihang University

Research projects in embodied AI and trustworthy agents

Research on agentic deception and safety alignment, visual long-document modeling, stochastic browser fingerprinting for risk-based authentication, and smart education agents.

Honors & Awards

  • National Scholarship (China; two-time recipient)
  • University Outstanding Student; Academic Excellence Scholarship
  • Merit Student (consecutive years); Outstanding Student Cadre; Outstanding Youth League Member
2026
Grand Prize and Second Prize, Capital Challenge Cup
2026
Honorable Mention, Mathematical Contest in Modeling (MCM)
2025
National Grand Prize (First Place), Challenge Cup
2025
National Bronze Award, China International College Students' Innovation Competition
2025
Second Prize, National College Mathematical Modeling Competition
2025
Team Third Prize, CCF “Wukong Cup” Quantum Programming Challenge
2023, 2024
First Prize, National College Mathematics Competition
2022
Provincial First Prize, National High School Mathematics League