About me
My name is Jiaxin Zhang, and I am a Senior Researcher of LLM & Agentic Game AI at Tencent, currently based in London, UK. I completed my PhD at the University of Strathclyde, where I was mentored by Dr. Yashar Moshfeghi. Before my doctoral studies, I obtained a Master’s degree from the University of Sheffield, under the supervision of Dr. Andreas Vlachos, following my Bachelor’s degree from Chang’an University. Additionally, I have gained professional experience as an NLP Algorithm Engineer at TCL Corporate Research Co., Ltd, where I worked for a year and a half.
My research is centered on advancing the numerical reasoning capabilities of Artificial Intelligence (AI) to bring them closer to human-level proficiency. More recently, my focus has shifted towards LLM-driven agentic AI for games, spanning both companion agents that interact naturally with human players and autonomous agents that adaptively play games on their own.
News
- [2026.03] Co-presented “AI Companion System in Peacekeeper Elite” at GDC 2026 (Game & Production Technology track), together with Bing Xue and Chen Wang from LightSpeed Studios, introducing the AI companion system in PUBG Mobile with command obedience, free chat, memory, and nurturing capabilities.
Current Projects
- PUBG Mobile (和平精英) AI Teammate · 2024.12 – 2026
- Lead of the player intent recognition model, responsible for translating in-game context and player instructions into actionable understanding for the AI companion.
- Focused on LLM-based decision making that drives the companion’s command obedience, memory, and interactive behaviors.
- Publicly introduced at GDC 2026.
- Auto-Gaming: Adaptive Game-Playing Agents
- Co-lead of the overall agent framework, and lead of the core module for language-directed agent action, translating high-level natural-language instructions into executable in-game actions.
- Exploring generalizable, LLM/agent-based decision-making that allows an agent to adaptively play and complete diverse games without game-specific engineering.
Publications
(for full publication list please checkout my Google Scholar or Semantic Scholar)
- Time is Not a Label: Continuous Phase Rotation for Temporal Knowledge Graphs and Agentic Memory
- Weixian Waylon Li, Jiaxin Zhang, Xianan Jim Yang, Tiejun Ma, Yiwen Guo
- EMNLP 2026 Main Conference — Oral · Corresponding author & research intern mentor
- RoMem models temporal evolution through continuous phase rotation and relation-dependent semantic speed.
- From System 1 to System 2: A Survey of Reasoning Large Language Models
- Duzhen Zhang, Zhong-Zhi Li, Ming-Liang Zhang, Jiaxin Zhang, et al.
- IEEE TPAMI, 48(3):3335–3354, 2026 (online 2025)
- CueTip: An Interactive and Explainable Physics-aware Pool Assistant
- Sean Memery, Kevin Denamganaï, Jiaxin Zhang, Zehai Tu, Yiwen Guo, Kartic Subr
- ACM SIGGRAPH 2025 Conference Papers · Research intern mentor
- Combines language-model interaction, physics simulation and expert rules for grounded shot advice and explanations.
- CMMaTH: A Chinese Multi-modal Math Skill Evaluation Benchmark for Foundation Models
- Zhongzhi Li, Ming-Liang Zhang, …, Jiaxin Zhang, Cheng-Lin Liu
- COLING 2025
- GeoEval: Benchmark for Evaluating LLMs and Multi-Modal Models on Geometry Problem-Solving
- Jiaxin Zhang, Zhong-Zhi Li, Ming-Liang Zhang, Fei Yin, Cheng-Lin Liu, Yashar Moshfeghi
- Findings of ACL 2024 · First author
- GOLD: Geometry Problem Solver with Natural Language Description
- Jiaxin Zhang, Yashar Moshfeghi
- Findings of NAACL 2024 · First author
- ELASTIC: Numerical Reasoning with Adaptive Symbolic Compiler
- Jiaxin Zhang, Yashar Moshfeghi
- NeurIPS 2022 · First author
For More Info
More info about please see my CV.
