Hi, I am Hengle Jiang. I am a PhD student at Department of Computer Science and Engineering, Southern University of Science and Technology, supervised by Prof. Ke Tang (IEEE Fellow). I obtained my B.Eng. degree from the same department, where I was fortunate to be mentored by Prof. Qi Hao and Prof. Dachuan Li. I am also a research intern at MINSys Group @ HKUST CSE, supervised by Prof. Xiaomin Ouyang.

I am always open to research collaborations. Feel free to reach out via email or view my work on Google Scholar.

Research Interests

My research focuses on building Safe, Robust, and Reliable autonomy systems that can operate effectively in complex, real-world environments.

Agent Safety: We investigate a new type of safety for LLM-based autonomous agents under non-adversarial settings. We introduce the concept of Agentic Pressure [ICLR’26, ACL’26] and find that agents sacrifice safety constraints to achieve task goals when placed under realistic operational pressure without any adversarial inputs. This work helps to bridge the gap between static benchmarks and dynamic real-world environments.
Agent Evaluation: We develop fine-grained, diagnostic assessment frameworks to better understand and quantify agent capabilities. MOAE is a self-evolving method that jointly optimizes agents for multiple objectives.

News

2026.09 🎉🎉 Our new paper is released on arXiv.
- MOAE: Multi-Objective Agent Evolution with Pareto-Preserving Search
2026.04 🎉🎉 One paper is accepted by ACL 2026 Findings.
- Why Agents Compromise Safety Under Pressure
2026.03 🏆🏆 Honored to be selected for the ACM SIGBED Student Travel Award ($1000) for CPS-IoT Week 2026!
2026.02 🎉🎉 One paper is accepted by ICLR 2026 Workshop Reliable Autonomy.
- Agentic Pressure: The Endogenous Entropy of Reliable Autonomy
2026.01 🎉🎉 Two papers are accepted by ACM/IEEE SenSys 2026.
- MoViD: View-Invariant 3D Human Pose Estimation via Motion-View Disentanglement
- WearBCI Dataset: Understanding and Benchmarking Real-World Wearable Brain-Computer Interfaces Signals
2025.09 🎉🎉 One paper is accepted by ACM MobiCom 2025 Demo.
- FreePose: Real-Time View-Invariant 3D Human Pose Estimation via Motion-View Disentanglement

Show older news

2025.06: 🏆🏆 I was awarded the Best Undergraduate Thesis Prize from SUSTech.

Selected Publications * Equal contribution

MOAE: Multi-Objective Agent Evolution with Pareto-Preserving Search

Paper

arXiv arXiv Preprint

Hengle JIANG, Qijun Cai, Ziying Luo, Ke Tang

MOAE formulates agent refinement as a Pareto-preserving evolutionary search over complete rollouts, jointly optimizing multiple objectives, including task performance, trajectory quality, and safety, without collapsing them into a fixed scalar objective.

Why Agents Compromise Safety Under Pressure

Paper

ACL'26 ACL 2026 Findings

Hengle JIANG, Ke Tang

Why benign agents fail without adversarial attacks? We found that LLM-based Agents exhibit normative drift under compounding task constraints in real-world applications. We provide detailed pressure taxonomy and a safety framework to mitigate such behavior.

MoViD: View-Invariant 3D Human Pose Estimation via Motion-View Disentanglement

Paper Code Video

SenSys'26 ACM/IEEE SenSys 2026 (51/271 accepted)

Yejia Liu^*, Hengle JIANG^*, Haoxian Liu, Runxi Huang, Xiaomin Ouyang

MoViD enables robust, cross-view 3D human pose estimation and real-time edge deployment by explicitly disentangling motion from view information and employing a view-aware inference strategy.

See all publications →

Educations

2025.09 - now, Doctor of Philosophy, Computer Science and Technology, SUSTech.
2021.09 - 2025.06, Bachelor of Engineering, Computer Science and Technology, SUSTech.

Internships

2024.08 - now, MINSys Group @ HKUST CSE, Hong Kong SAR, China.
2022.06 - 2023.04, SZ DJI Technology Co.,Ltd.

Teaching

2026 Spring, CS311H: Artificial Intelligence (Honor Track), SUSTech, Teaching Assistant
2025 Fall, CS112: Introduction to Python Programming, SUSTech, Teaching Assistant

Services

ICLR 2026 Workshop Reliable Autonomy, Reviewer