Leyi Pan (潘乐怡)

About Me

Hi! My name is Leyi Pan (潘乐怡). I am currently a 3rd-year Ph.D. student at Tsinghua University. Before that, I received my B.Eng. from Tsinghua University in 2024.

I am currently a member of the Top Talent Intern Program at Minimax, working in the Foundation Model Team, Reinforcement Learning Group, where I contribute to releases of the M-series LLMs. From July 2025 to July 2026, I was a research intern on LLM reasoning and reinforcement learning at Tongyi Lab, Alibaba Group. Previously, I interned at Z.ai, where I contributed to the GLM-4.5V and GLM-4.1V-Thinking models.

My research interests mainly lie in:

LLM Reinforcement Learning (current focus)
Trustworthy AI

Feel free to reach out if you would like to discuss research or explore potential collaboration!

News

2026.07 Joined Minimax through the Top Talent Intern Program.
2026.07 Omni-SafetyBench is accepted by ACM MM 2026!
2026.07 MarkDiffusion is accepted by JMLR!
2026.06 MarkLLM has reached 1k stars!
2026.04 d-TreeRPO is accepted by ACL 2026 Main!
2025.07 Joined Tongyi Lab, Alibaba Group, as a Research Intern.
2025.07 GLM-4.5V and GLM-4.1V-Thinking are released on arXiv! (76 authors including Leyi Pan)
2025.05 Can LLM Watermarks Robustly Prevent Unauthorized Knowledge Distillation? is accepted by ACL 2025 Main.
2024.10 Joined Z.ai AI Lab as a Research Intern.

Education

2024.09 - Present, Ph.D. @ Tsinghua University, Beijing, China.
2020.09 - 2024.06, B.E. @ Tsinghua University, Beijing, China.

Internships

2026.07 - PresentMinimax, Top Talent Intern Program
Foundation Model Team, Reinforcement Learning Group. Contributing to releases of the M-series LLMs.
2025.07 - 2026.07Tongyi Lab, Alibaba Group, Research Intern
Topic: LLM reasoning and reinforcement learning.
2024.10 - 2025.06AI Lab, Z.ai, Research Intern
Topic: Reasoning and reinforcement learning for multimodal large language models.
2024.06 - 2024.08CUHK MISC Lab, Research Assistant
Topic: Trustworthy AI.
2023.06 - 2023.08KuaiShou Technology, Research Intern
Topic: Tool-integrated agent.

Projects

GLM-4.5V (Contributor)

A versatile multimodal reasoning model with strong visual understanding, coding, STEM, and GUI-agent capabilities.

Paper | Code

| Hugging Face | ModelScope

GLM-4.1V-Thinking (Contributor)

An open-source multimodal reasoning model trained with scalable reinforcement learning for challenging visual and language tasks.

Paper | Code

| Hugging Face | ModelScope

MarkLLM (Project Lead & First Contributor)

An open-source toolkit for LLM watermarking, covering watermark generation, detection, evaluation, visualization, and extensible algorithm implementations.

Paper | Code

MarkDiffusion (Project Lead & First Contributor)

An open-source toolkit for generative watermarking of latent diffusion models, designed for reproducible evaluation and easy integration of watermarking methods.

Paper | Code

Selected Publications & Preprints

A full publication list is available on my Google Scholar page. (* denotes equal contribution)

LLM Reinforcement Learning (current focus)

RLCSD: Reinforcement Learning with Contrastive On-Policy Self-Distillation

Leyi Pan, Shuchang Tao, Yunpeng Zhai, Lingzhe Zhang, Zhaoyang Liu, Bolin Ding, Aiwei Liu, Lijie Wen.

Preprint, 2026 | PDF | Code
d-TreeRPO: Towards More Reliable Policy Optimization for Diffusion Language Models

Leyi Pan, Shuchang Tao, Yunpeng Zhai, Zheyu Fu, Liancheng Fang, Minghua He, Lingzhe Zhang, Zhaoyang Liu, Bolin Ding, Aiwei Liu, Lijie Wen.

ACL Main, 2026 · CCF-A | PDF | Code
GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning

Zhipu AI (76 authors including Leyi Pan).

Technical Report, 2025 | PDF | Code

Trustworthy AI

Omni-SafetyBench: A Benchmark for Safety Evaluation of Audio-Visual Large Language Models

Leyi Pan, Zheyu Fu, Yunpeng Zhai, Shuchang Tao, Sheng Guan, Shiyu Huang, Lingzhe Zhang, Zhaoyang Liu, Bolin Ding, Felix Henry, Aiwei Liu, Lijie Wen.

ACM MM, 2026 · CCF-A | PDF | Code
MarkDiffusion: An Open-Source Toolkit for Generative Watermarking of Latent Diffusion Models

Leyi Pan, Sheng Guan, Zheyu Fu, Luyang Si, Huan Wang, Zian Wang, Hanqian Li, Xuming Hu, Irwin King, Philip S. Yu, Aiwei Liu, Lijie Wen.

JMLR · CCF-A | PDF | Code
Can LLM Watermarks Robustly Prevent Unauthorized Knowledge Distillation?

Leyi Pan, Aiwei Liu, Shiyu Huang, Yijian Lu, Xuming Hu, Lijie Wen, Irwin King, Philip S. Yu.

ACL Main, 2025 · CCF-A | PDF | Code
WaterSeeker: Pioneering Efficient Detection of Watermarked Segments in Large Documents

Leyi Pan, Aiwei Liu, Yijian Lu, Zitian Gao, Yichen Di, Lijie Wen, Irwin King, Philip S. Yu.

NAACL Findings, 2025 | PDF | Code
MarkLLM: An Open-Source Toolkit for LLM Watermarking

Leyi Pan, Aiwei Liu, Zhiwei He, Zitian Gao, Xuandong Zhao, Yijian Lu, Binglin Zhou, Shuliang Liu, Xuming Hu, Lijie Wen, Irwin King, Philip S. Yu.

EMNLP Demo, 2024 | PDF | Code
A Survey of Text Watermarking in the Era of Large Language Models

Aiwei Liu*, Leyi Pan*, Yijian Lu, Jingjing Li, Xuming Hu, Lijie Wen, Irwin King, Philip S. Yu.

ACM Computing Surveys | PDF
A Semantic Invariant Robust Watermark for Large Language Models

Aiwei Liu, Leyi Pan, Xuming Hu, Shiao Meng, Lijie Wen.

ICLR, 2024 · CCF-A | PDF | Code
An Unforgeable Publicly Verifiable Watermark for Large Language Models

Aiwei Liu, Leyi Pan, Xuming Hu, Shuang Li, Lijie Wen, Irwin King, Philip S. Yu.

ICLR, 2024 · CCF-A | PDF | Code