MeshHeal: Two-Timescale Self-Healing for Gray Failures in Decentralized LLM Agent Networks
TL;DR Peer review catches unreliable agent outputs, while longer-term monitoring guides exclusion and recovery.
Reinforcement learning · Language models
About me
I am a PhD student in Electrical Engineering at Arizona State University, advised by Prof. Shaofeng Zou. I also work closely with Prof. Sen Lin and Prof. Yingbin Liang. Previously, I received my B.Eng. in Automation from Xi’an Jiaotong University.
I am interested in reinforcement learning, LLM post-training, and AI agents (including multi-agent systems), with a focus on safety, alignment, and decision-making under constraints. I am also interested in applications of these methods to healthcare.
News
Since Aug 2025- My paper Information-Directed Offline-to-Online Reinforcement Learning was accepted to NeurIPS 2026!
- New preprint: MeshHeal, on detecting and recovering from degraded performance in LLM agent networks.
- Our paper HIPO was accepted to EMNLP 2026 (Main Conference)!
- Information-Directed Offline-to-Online Reinforcement Learning appeared at the ICML 2026 DEMO Workshop.
- Our paper Marvel, on safe offline-to-online reinforcement learning, was accepted to TMLR!
- Started my PhD at Arizona State University, advised by Prof. Shaofeng Zou.
Publications
Google ScholarInformation-Directed Offline-to-Online Reinforcement Learning
Main Conference · Also at ICML 2026 DEMO Workshop
TL;DR Direct online exploration toward the uncertainty that remains after learning from offline data.
HIPO: Instruction Hierarchy via Constrained Reinforcement Learning
Main Conference
TL;DR Train language models to help users while treating system-instruction compliance as an explicit constraint.
Towards Fast Safe Online Reinforcement Learning via Policy Finetuning
TL;DR Align value estimates and adapt constraint penalties to safely fine-tune offline policies online.
Fourier Ptychographic Microscopy 10 Years On: A Review
13(4), 324
TL;DR A review of computational microscopy, from image reconstruction to digital pathology and deep learning.
Education
2025–present
Arizona State University
PhD in Electrical Engineering · Advisor: Shaofeng Zou
2021–2025
Xi’an Jiaotong University
B.Eng. in Automation
Service
Reviewer
NeurIPS 2026 ICLR 2027 Scientific Reports IEEE/ACM Transactions on Networking
Photography
View galleryAway from my desk, I like to photograph landscapes, cities, and the night sky.
A few recent favorites. Click a photograph to take a closer look.