Leilei Gan (甘磊磊), Ph.D.

Email: [email protected]

Address: 38 Zheda Road, Hangzhou, Zhejiang, China

About Me

I am a tenure-track Assistant Professor (Ph.D. advisor) at the School of Software Technology, Zhejiang University. I received my Ph.D. in Computer Science from Zhejiang University in March 2023, advised by Prof. Fei Wu and Prof. Yi Yang. From March to September 2023, I was a visiting Ph.D. student at Nanyang Technological University, hosted by Prof. Luu Anh Tuan.

My current research focuses on foundation models, agents, reinforcement learning, mechanistic interpretability ****and AI applications in law, finance, and industry. I have published nearly 30 papers in leading conferences and journals, including IEEE T-PAMI, ICLR, NeurIPS, ACL, and AAAI.

My research has been supported by the National Natural Science Foundation of China (NSFC), the "Pioneer & Leading Goose" R&D Program of Zhejiang, the CCF-Taobao & Tmall Research Fund, the CCF-Ant Group Research Fund, the CCF-1688 Research Fund, and the CAAI-MindSpore Open Research Fund.

I was selected for the Youth Science and Technology Innovation Leading Talent Program of Ningbo. My work received the Wen-Tsun Wu AI Science and Technology Progress Award (First Prize, 2024) and the Zhejiang Provincial Science and Technology Progress Award (First Prize, 2024).

News

<aside> 🚨

Ph.D. Student Opening — Deadline: September 8, 2026

I am currently recruiting direct-entry Ph.D. students to work on large language models, AI agents, and reinforcement learning. Interested students are encouraged to send their CV to [email protected] by September 8, 2026.

</aside>

• 2026.08 Three papers were accepted to EMNLP 2026 (two Main Conference papers and one Findings paper).
• 2026.06 Our survey on Direct Preference Optimization was accepted to IEEE T-PAMI.
• 2026.06 The undergraduate thesis Deep Reasoning of Multimodal Large Models via Reinforcement Learning, completed by Shuang Chen under my supervision, was selected as an Outstanding Undergraduate Thesis by the College of Computer Science and Technology, Zhejiang University.
• 2026.04 Four papers were accepted to the ACL 2026 Main Conference.
• 2025.12 Our work on agent environment tuning was accepted to ICLR 2026.
• 2025.10 Our work on fast-slow thinking GRPO was accepted to NeurIPS 2025 as a Spotlight (Top 3.2%).
• 2025.08 One paper was accepted to Findings of EMNLP 2025.
• 2025.05 T2I-FactualBench was accepted to the ACL 2025 Main Conference.
• 2024.12 Two papers were accepted to AAAI 2025.
• 2024.11 Legal-HalBench was accepted to COLING 2025 as an Oral paper.

📚 Publications

(† = corresponding author; ∗ = equal contribution. Full list on Google Scholar.)

2026

1. Zongyue Li, Chengyue Yu, Lei Zang, Chenyi Zhuang, Linjian Mo, Leilei Gan†. Last Step Matters: Early Uncertainty Cannot Predict Failure in Long-Horizon Agents. EMNLP 2026