Jiashen Liu

Embodied AI · Visual Perception · Agentic Systems

Hi! I’m Jiashen Liu (刘嘉申), a researcher passionate about building intelligent systems that can perceive, reason, and act in the real world.

My research sits at the intersection of embodied AI, computer vision, and agentic systems. I’m particularly interested in:

🤖 Embodied AI & Robotics
Developing perception systems for challenging environments — from underwater caves to complex 3D spaces. I work on visual-inertial navigation, simulation-driven adaptation, and multi-sensor fusion for robust autonomous systems.

🎥 Video & Vision-Language Understanding
Exploring how geometric perception and spatial reasoning can enhance video-language models. How do AI systems truly “see” video beyond frame sequences?

🧠 LLM Agents & Tool Use
Building reliable agentic systems with long-horizon planning, self-evolution capabilities, and intelligent tool orchestration. From automated security testing to multimodal retrieval.

🎓 Education

MSc in Cyber Security Engineering (with Distinction)
University of Warwick, UK | 2024 - 2025
Thesis: WebPentestRAG - Retrieval-Augmented Generation for Automated Web Application Penetration Testing

BM in Information Management and Information Systems
Nanjing Agricultural University, China | 2020 - 2024
Average Grade: 84.25/100 | Outstanding Graduation Thesis Award

Summer Program - Technology, Science and Environment
The Chinese University of Hong Kong, Hong Kong | 2023

💡 Recent Highlights

  • 🏆 Global Excellence Award at GeoMind-OreSense Entrepreneur Dreamer Competition (2026.04)
  • 📝 Multiple papers accepted/submitted to top-tier conferences including ICARCV 2026, IEEE TMC, ICRA 2027
  • 🔬 Research experience spanning underwater robotics, LLM agents, and multimodal AI systems

I’m always excited to discuss research ideas and potential collaborations. Feel free to reach out!

Jiashen Liu's profile avatar
Email Google Scholar GitHub

Publications & Manuscripts

Published / Accepted

  1. FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge Devices

    C. Li , S. Huang , J. Liu , M. Lei , J. Xing , K. Wu , L. Wang , F. Luo

    IEEE Transactions on Mobile Computing · Accepted · 2026

    arXiv PDF

  2. Active Learning-Driven Ensemble Framework for Efficient and Robust Reliability Analysis in Geotechnical Engineering

    X. Liu , B. Liu , Y. Liu , J. Liu , A. Wang

    Machine Learning and Data Science in Geotechnics, 2(1), 55–66 · 2026

    Paper

  3. SynUW-VIBench: A Synthetic Underwater Visual-Inertial Benchmark with Frame and Event Cameras

    X. Yu , J. Liu , et al.

    ICARCV 2026 · Oral Presentation · 2026

    X. Yu and J. Liu contributed equally.

Preprints / Under Review

  1. CAVERN: Controllable and Validated Environments for Robotic Navigation in Underwater Caves

    J. Liu , J. Zhou , K. Zhang , X. Yu , Z. Hong

    Submitted to ICRA 2027 · 2026

    Corresponding author: Z. Hong.

  2. PI-TTA: Physics-Informed Source-Free Test-Time Adaptation for Robust Human Activity Recognition on Mobile Devices

    C. Li , L. Wang , M. Lei , J. Liu , Y. Zhang , K. Wu , F. Luo

    Preprint · Submitted to IEEE Transactions on Mobile Computing · 2026

    arXiv PDF

In Preparation

  1. How Do VLMs Actually See Video? A Geometric Account of Visual Context Representations in Video–LLMs

    J. Liu , S. Xia

    Manuscript in preparation for ICLR 2027 · 2026