Released two new studies on evaluation design and action settlement in intelligent-agent environments.
LLM Agents · Reliable AI · Computer Vision
Haotian Chen
陈皓天M.S. student in Cyberspace Security at the University of Science and Technology of China.
I study how learning-based agents can perceive, remember, and act reliably in open-ended environments. My work develops evaluation protocols and system mechanisms for auditable action execution, persistent memory, and grounded multimodal reasoning.
Previously, I completed a dual-degree undergraduate programme in Computer Science at the Southwest Jiaotong University–Leeds Joint School, with degrees awarded by Southwest Jiaotong University and the University of Leeds. I also conducted research at Tsinghua AIR and worked with the operating-system product team at ZTE.
01
News
Began my M.S. in Cyberspace Security at the University of Science and Technology of China.
Graduated with First-Class Honours from the University of Leeds and a B.Eng. from Southwest Jiaotong University, and was recognized as an Outstanding Graduate.
DPNet was published in IEEE Internet of Things Journal.
02
Research
My research asks how intelligent agents can remain reliable, auditable, and adaptive when decisions unfold over long horizons, under partial observability, and through interaction with users and environments.
Reliable & Auditable LLM Agents
Evaluation and control for tool-using and multi-agent systems, with an emphasis on action semantics, order sensitivity, progress attribution, replayability, and failure diagnosis.
Long-Term Memory & Personalization
Mechanisms for agents to acquire, update, retrieve, and forget long-term memories while preserving user preferences, temporal consistency, and controllable behavior.
Multi-Agent & Social Simulation
Executable environments for studying coordination, interaction, and emergent behavior through traceable world-state transitions and reproducible counterfactual experiments.
Embodied & 3D Intelligence
Multimodal models that connect vision, language, and geometry for tiny-object perception, pose understanding, 3D scene reasoning, and embodied decision-making.
03
Selected Publications
When Does Exercise-Specific Joint Selection Help? An Audit of Evaluation and Control Design
An evaluation audit of skeleton-based exercise correctness classification, separating estimands, subset structure, and temporal controls.
Auditing Action Settlement in LLM Agent Environments: Order, Progress, and Replay
A typed settlement contract and audit framework for order sensitivity, useful progress, and replay consistency in multi-agent environments.
DPNet: Dynamic Pooling Network for Tiny Object Detection
Input-aware dynamic downsampling for more efficient tiny-object detection in unmanned aerial imagery.
Dance of Fireworks: An Interactive Broadcast Gymnastics Training System Based on Pose Estimation
A mobile pose-estimation system that links exercise feedback with responsive visual rewards.
04
Education & Experience
University of Science and Technology of China
M.S. in Cyberspace Security, focusing on large language model security and reliable agents.
Southwest Jiaotong University
Undergraduate study in Computer Science and Technology at the SWJTU–Leeds Joint School. Average 91.23; ranked 8 / 82.
University of Leeds
Dual-degree undergraduate programme in Computer Science through the SWJTU–Leeds Joint School.
Tsinghua AIR · Research Intern
Collaborated on computer vision and AI research for robot target recognition and autonomous navigation.
ZTE Chengdu · Software Development Intern
Built an automated kernel-configuration inspection tool, reducing a 30-minute review to about 5 minutes with over 98% accuracy.