Hi there Welcome to my Homepage!

I am an undergraduate (2023–2027) at Southwest Jiaotong University. My research interests include reinforcement learning, agentic LLMs, and social recommendation.

I work at Cityu with Prof. Songhua Hu. Previously I worked at ETS lab@THU with Prof.Yang Liu Currently I conduct the Agent memory research at ailab.

News

• A2GBD is accepted in WWW 2026 🔥
• Our work achieved 7th place globally in the lightweight deployment track of Tencent 2025 OpenMind’s Honor of Kings competition. 🔥
• *DwT is accepted in MM 2025
• In charge of Cybersecurity Club. Feel free to reach out if you’d like to join.

Experience

City University of Hong Kong
Jun 2025 – Present
Research Intern, advised by Prof. Songhua Hu

Shanghai AI Lab
Dec 2025 – Present
Research Intern at AI45 Lab

ETS Lab, Tsinghua University
Jan 2025 – Dec 2025
Research Intern, working on LLMs for Transportation and Mobility

Southwest Jiaotong University
Jan 2024 – Jul 2027
Academic Innovation Scholarship
B.E. in Transportation Engineering at School of Transportation and Logistics, advised by Prof. Tianrui Li

Publications

A2GBD: Attack-Agnostic Graph Backdoor
Chenxu Du, Yang Liu†
We propose Attack-Agnostic Graph Backdoor (A2GBD), a unified backdoor defense framework for reinforcement learning.
WWW 2026   [arXiv] [code]

Draw with Thought: Unleashing Multimodal Reasoning for Scientific Diagram Generation
We propose DwT, a training-free framework that unleashes multimodal reasoning for scientific diagram generation by guiding MLLMs to reconstruct diagrams.
ACM MM 2025   [arXiv]

Projects

RL Lab
RL Lab is dedicated to developing efficient and adaptable deep reinforcement learning algorithms.
[code]

MLLM-Reasoning-Enhancement-Guide
This repository offers a systematic guide to MLLM reasoning enhancement, including method overviews, key papers, tools, datasets, and tutorial notes
[website]

Awards ——–

• Academic Innovation Scholarship 2025

Talks