Wenxuan Wang | HIT SCIR

About Me

I am a Joint PhD Student at Shanghai AI Lab and HIT-SCIR, advised by Yu Cheng and Weinan Zhang.

My research interests lie in AI agents and generative models. My work explores memory and planning in long-horizon agents, as well as transferring agent experience into generative models. I am interested in how agents can learn from experience and how agents and generative models can improve together.

Research Interests

AI Agents: Memory, planning, and learning from experience for long-horizon tasks.
Generative Models: Transferring agent experience into generative models and exploring their co-evolution.

News

[Aug. 2026] Our paper SubtleMemory has been accepted to EMNLP 2026 Findings!
[Aug. 2026] Our paper Agent Planning Benchmark (APB) has been accepted to EMNLP 2026 Findings!

Publications

arXiv

Internalizing Agent Experience into Diffusion Model Weights via On-Policy Context Distillation

Wenxuan Wang^*, Zekai Liu^*, Weinan Zhang, Yu Cheng^†, Yang Yang^†

arXiv, 2026.

PDF Code Project Page BibTeX
EMNLP Findings

SubtleMemory: A Benchmark for Fine-Grained Relational Memory Discrimination in Long-Horizon AI Agents

Wenxuan Wang^*, Haoyu Sun^*, Fukuan Hou, Mingyang Song, Weinan Zhang, Yu Cheng^†, Yang Yang^†

EMNLP 2026 Findings

PDF Code Project Page BibTeX
EMNLP Findings

Agent Planning Benchmark: A Diagnostic Framework for Planning Capabilities in LLM Agents

Haoyu Sun^*, Wenxuan Wang^*, Mingyang Song, Jujie He, Weinan Zhang, Yang Liu, Yang Yang^†, Yu Cheng^†

EMNLP 2026 Findings

PDF Code BibTeX

Selected Projects

September 2024 – January 2025

A multi-agent sandbox for simulating social media interactions and studying agent behavior.

Contributed to system iteration and performance optimization, improving the efficiency of multi-agent simulations.

March 2025 – May 2025

An autonomous social media agent system for exploring agent autonomy and influence in online interactions.

Developed the central control module and supported real-world testing and iterative refinement of the system.

Education

Aug. 2026 – Present
Harbin Institute of Technology
Ph.D. Student, Faculty of Computing
Joint training with Shanghai AI Laboratory
Aug. 2022 – Jun. 2026
Harbin Institute of Technology
Bachelor's degree in Artificial Intelligence
School of Future Technology

Internship Experience

Oct. 2025 – Oct. 2026
Shanghai AI Laboratory
Intern
Researched AI agent memory, planning, and experience distillation, resulting in three co-first-authored papers: two accepted to Findings of EMNLP 2026 and one manuscript under review.

Selected Honors & Awards

May 2026
Outstanding Graduate, Harbin Institute of Technology
Mar. 2026
Song Qingling Grant
May 2025
Second Prize, ASC Student Supercomputer Challenge
Mar. 2025
Samsung Scholarship
Mar. 2025
Song Qingling Grant
Nov. 2023
First Prize, National College Student Mathematics Competition (Heilongjiang Division)