Haoran Xu

About

I am a PhD student in the James Watt School of Engineering at the University of Glasgow.

My research focuses on trustworthy multi-agent AI systems, with a particular interest in how collaborative AI agents can decide when to commit, when to abstain, and when to safely abort under adversarial disagreement, semantic poisoning, and unreliable evidence.

More broadly, I work on protocol-level mechanisms that connect ideas from distributed computing, Byzantine resilience, robust aggregation, and evidence-aware decision making to improve the reliability of LLM-agent collaboration.

Research agenda

My current research agenda is to build trustworthy control and commitment mechanisms for multi-agent AI systems. I am particularly interested in three questions:

1. When should a multi-agent system commit? Designing mechanisms that distinguish reliable agreement from superficial or unsafe consensus.
2. How can agent collaboration remain robust under semantic attacks? Studying Byzantine-style adversarial behaviors in natural-language proposals, evidence use, and agent reasoning.
3. How can commitment be certified? Developing certificate-based mechanisms that make agent decisions auditable, reproducible, and safe to reject when agreement is insufficient.

Selected work

• Hierarchical Certified Semantic Commitment for Trustworthy Multi-Agent Collaboration
  Haoran Xu, Lei Zhang, Iadh Ounis, Xianbin Wang
  arXiv preprint, 2026
  paper coming soon · code coming soon
• A Multi-Round Voting Consensus Framework via Evolutionary Game Theory and Score Completion
  Haoran Xu, Zhongxu Dong, Yixuan Fan, Huanyu Wu, Qiming Li, Yao Sun, Lei Zhang
  IEEE International Conference on Communications Workshops (ICC Workshops), 2026
  paper · code · slides · bibtex coming soon

Selected projects

• H-CSC
  Hierarchical Certified Semantic Commitment for trustworthy multi-agent collaboration.
  code coming soon
• CRSE / FBA Encoder
  A robust semantic encoder for Byzantine-resilient semantic agreement.
  github.com/HrxuAlbert/FBA_ENCODER
• LLM API Demos
  Lightweight demos for structured LLM API usage, agent workflows, and research prototyping.
  github.com/HrxuAlbert/LLM-API-demos
• Work Planner
  A local-first academic workflow tool for research planning and execution tracking.
  github.com/HrxuAlbert/Work-Planner

Education

University of Glasgow
PhD in Engineering, James Watt School of Engineering, 2024–present