
My name in Chinese:
邹安妮
Email:
[email protected]
<aside>
<img src="notion://custom_emoji/0b016447-bbab-499d-8778-4b0b95ff6c15/40889fc3-57f5-81cf-90f6-007a4e1d77fa" alt="notion://custom_emoji/0b016447-bbab-499d-8778-4b0b95ff6c15/40889fc3-57f5-81cf-90f6-007a4e1d77fa" width="40px" />
Google Scholar
</aside>
<aside>
<img src="notion://custom_emoji/0b016447-bbab-499d-8778-4b0b95ff6c15/40889fc3-57f5-81a4-92e5-007a8be8ee99" alt="notion://custom_emoji/0b016447-bbab-499d-8778-4b0b95ff6c15/40889fc3-57f5-81a4-92e5-007a8be8ee99" width="40px" />
GitHub
</aside>
👋 Hi!
I am an Algorithm Engineer at Alibaba Token Foundry (Tongyi Lab), working on LLM agents and post-training techniques. I received my Master’s and B.S. degree at Shanghai Jiao Tong University, under the supervision of Prof. Hai Zhao, where I also collaborated closely with Dr. Zhuosheng Zhang.
My research centers on learning and evolving agentic intelligence through interaction. I study how agents turn experience and feedback into generalizable capabilities, from efficient self-improvement (AgentEvolver) to learning across diverse, interactive environments (Cowork-Gym). My long-term goal is to build agents that continually learn, adapt, and become useful participants in everyday life.
<aside>
🌱 What’s New
- 10/2026: Introduced Cowork-Gym, a scalable framework that synthesizes interactive work environments for agent post-training. By combining executable workspaces, stateful user simulation, and verifiable outcomes, it lets agents learn to use tools, work with files, and adapt to changing user requirements. [**Paper][**Blog]
- 04/2026: PawFriends is now live, offering a shared social space where autonomous agents initiate activities and interact with one another. [**Demo]**
- 01/2026: Released AgentEvolver Game Arena, a multi-agent social game platform featuring Avalon and Diplomacy, designed to study models’ social reasoning capabilities and develop efficient reinforcement learning methods for ultra-long context scenarios. [**GitHub]**
- 11/2025: Released AgentEvolver, a self-evolving agent learning system built around self-questioning, self-navigating, and self-attributing. [**Paper][**GitHub]
</aside>
📚 Publications
First / Co-first Author
- Cowork-Gym: Synthesizing Interactive Work Environments at Scale
A Zou*, C Chen*, J Deng*, W Zhou*, X Mai*, Y Wu*, J Lin, L Xie, L Qin, Q Zhao, et al.
Paper Cited 1
- AgentEvolver: Towards Efficient Self-Evolving Agent System
Y Zhai*, S Tao*, C Chen*, A Zou*, Z Chen*, Q Fu*, S Mai*, L Yu, J Deng, Z Cao, et al.
Paper Cited 115 GitHub ★1.6k
- DocBench: A Benchmark for Evaluating LLM-based Document Reading Systems
A Zou, W Yu, H Zhang, K Ma, D Cai, Z Zhang, H Zhao, D Yu
Paper Cited 73 GitHub ★90
- MedAgents: Large Language Models as Collaborators for Zero-shot Medical Reasoning
X Tang*, A Zou*, Z Zhang, Z Li, Y Zhao, X Zhang, A Cohan, M Gerstein
Paper Cited 659 GitHub ★373
- AuRoRA: A One-for-all Platform for Augmented Reasoning and Refining with Task-Adaptive Chain-of-Thought Prompting
A Zou, Z Zhang, H Zhao
Paper Website Cited 3 GitHub ★6
- Generalizable Chain-of-Thought Prompting in Mixed-task Scenarios with Large Language Models
A Zou, Z Zhang, H Zhao, X Tang
Paper Cited 22 GitHub ★99
- Decker: Double Check with Heterogeneous Knowledge for Commonsense Fact Verification
A Zou, Z Zhang, H Zhao
Paper Cited 26 GitHub ★11
- EventBERT: Incorporating Event-based Semantics for Natural Language Understanding
A Zou, Z Zhang, H Zhao
Paper Cited 1
Collaborative Publications
<aside>
💼 Experience
Alibaba Token Foundry (Tongyi Lab)
Algorithm Engineer · 2025.05–Present
Meituan, Core Algorithms Group
Algorithm Intern · 2024.07–2024.10
Tencent AI Lab
Research Intern · 2024.01–2024.07 · Advised by Wenhao Yu
</aside>
<aside>
🎓 Education
Shanghai Jiao Tong University · Master’s, NLP · 2022.09–2025.03 · Advised by Hai Zhao | Collaborated with Zhuosheng Zhang
Shanghai Jiao Tong University · Bachelor’s, Information Engineering · 2018.09–2022.06
</aside>
🏅 Honors and Awards
- National Scholarship, 2022–2023
- Class A Scholarship (top 5%)
- Outstanding Graduate, SJTU
- Junyuan Scholarship for 4 consecutive years