Peng Li
Research Associate Professor
About
Peng Li is a Research Associate Professor at the Institute for AI Industry Research (AIR), Tsinghua University. Prior to joining Tsinghua, he served as a Principal Researcher and Team Leader at WeChat AI, Tencent, and previously worked at the Institute of Deep Learning (IDL), Baidu. His research interests include Large Language Models (LLMs), LLM Agents, AI for Mathematics (AI4Math), AI Scientist, and Multimodal Large Language Models (MLLMs). He has received several research awards, including the ACL 2023 Outstanding Paper Award, the ICAIS 2025 Responsible AI Research Award, and a CIKM 2021 Best Resource Paper Nomination. He has led several major research projects, including a key task under the National Science and Technology Innovation - Major Program. He has also served as an Area Chair for top-tier international conferences such as ICLR, ACL, EMNLP, and NAACL. His research has been deployed in large-scale production systems at WeChat and Baidu, supporting tens of millions of users daily. He is also a recipient of the First Prize of the Qian Weichang Chinese Information Processing Science and Technology Award from the Chinese Information Processing Society of China (CIPS).
LLMs
LLM Agents
AI4Math
AI Scientist
MLLMs
Address: 11/F, Block C, Qidi Science & Technology Building, Tsinghua Science Park, Haidian District, Beijing
My current research focuses on AI4Math, AI Scientist, and LLM agents, with broader interests in reasoning and multimodal intelligence.
AI Mathematician: Towards Fully Automated Frontier Mathematical Research
Yuanhang Liu, Yanxing Huang, Yanqiao Wang, Peng Li, Yang Liu. Preprint. [arXiv] [System] [Slides] [Slides (in Chinese)] [Blog]
Shows how AI Mathematician can help tackle real research problems, moving AI for math beyond solving benchmark puzzles toward assisting frontier mathematical discovery.
AI Mathematician as a Partner in Advancing Mathematical Discovery - A Case Study in Homogenization Theory
ICAIS 2025 Responsible AI Research Award
A concrete case study showing AI as a research partner for mathematicians: it helps explore proof ideas while experts keep the final reasoning rigorous and trustworthy.
Pessimistic Verification for Open Ended Math Questions
Yanxing Huang, Zihan Tang, Zejin Lin, Peng Li, Yang Liu. ICML 2026. [arXiv]
Addresses a key obstacle for AI-assisted math: not just producing proofs, but reliably catching mistakes in open-ended mathematical reasoning.
FormaRL: Enhancing Autoformalization with no Labeled Data
Makes formal mathematics more accessible to AI systems by reducing the need for expensive labeled data when translating informal math into machine-checkable proofs.
AIGS: Generating Science from AI-Powered Automated Falsification
Zijun Liu, Kaiming Liu, Yiqi Zhu, Xuanyu Lei, Zonghan Yang, Zhenhe Zhang, Peng Li, Yang Liu. Preprint. [arXiv] [Project Webpage] [Slides (in Chinese)]
Explores what it would take for AI to contribute to science by emphasizing testing and falsification, so generated ideas are challenged rather than simply accepted.
Position: Towards Unified Alignment Between Agents, Humans, and Environment
Zonghan Yang, An Liu, Zijun Liu, Kaiming Liu, Fangzhou Xiong, Yile Wang, Zeyuan Yang, Qingyuan Hu, Xinrui Chen, Zhenhe Zhang, Fuwen Luo, Zhicheng Guo, Peng Li, Yang Liu. ICML 2024, 56251-56275. [pdf] [arXiv] [Project Webpage]
Broadens AI alignment from following human preferences to building agents that also respect the surrounding environment and practical constraints like time and cost.
A Dynamic LLM-Powered Agent Network for Task-Oriented Agent Collaboration
Shows that teams of AI agents can become more effective when they adapt who collaborates with whom instead of using a fixed discussion structure for every task.
AI4Math and AI Scientist
AI Mathematician: Towards Fully Automated Frontier Mathematical Research
Yuanhang Liu, Yanxing Huang, Yanqiao Wang, Peng Li, Yang Liu. Preprint. [arXiv] [System] [Slides] [Slides (in Chinese)] [Blog]
AI Mathematician as a Partner in Advancing Mathematical Discovery - A Case Study in Homogenization Theory
ICAIS 2025 Responsible AI Research Award
Pessimistic Verification for Open Ended Math Questions
Yanxing Huang, Zihan Tang, Zejin Lin, Peng Li, Yang Liu. ICML 2026. [arXiv]
FormaRL: Enhancing Autoformalization with no Labeled Data
AIGS: Generating Science from AI-Powered Automated Falsification
Zijun Liu, Kaiming Liu, Yiqi Zhu, Xuanyu Lei, Zonghan Yang, Zhenhe Zhang, Peng Li, Yang Liu. Preprint. [arXiv] [Project Webpage] [Slides (in Chinese)]
LLM Agents
Position: Towards Unified Alignment Between Agents, Humans, and Environment
Zonghan Yang, An Liu, Zijun Liu, Kaiming Liu, Fangzhou Xiong, Yile Wang, Zeyuan Yang, Qingyuan Hu, Xinrui Chen, Zhenhe Zhang, Fuwen Luo, Zhicheng Guo, Peng Li, Yang Liu. ICML 2024, 56251-56275. [pdf] [arXiv] [Project Webpage]
A Dynamic LLM-Powered Agent Network for Task-Oriented Agent Collaboration
ReAct Meets ActRe: Autonomous Annotation of Agent Trajectories for Contrastive Self-Training
Advancing Language Multi-Agent Learning with Credit Re-Assignment for Interactive Environment Generalization
Zhitao He, Zijun Liu, Peng Li, Yi R Fung, Ming Yan, Ji Zhang, Fei Huang, Yang Liu. COLM 2025. [pdf] [arXiv] [code] [Slides (in Chinese)]
Scaling External Knowledge Input Beyond Context Windows of LLMs via Multi-Agent Collaboration
Reasoning and Learning
Inference-Time Scaling for Generalist Reward Modeling
Zijun Liu, Peiyi Wang, Runxin Xu, Shirong Ma, Chong Ruan, Peng Li, Yang Liu, Yu Wu. Preprint. [arXiv]
Writing-RL: Advancing Long-form Writing via Adaptive Curriculum Reinforcement Learning
MUSEG: Reinforcing Video Temporal Understanding via Timestamp-Aware Multi-Segment Grounding
Enabling Stroke-Level Structural Analysis of Hieroglyphic Scripts without Language-Specific Priors
Fuwen Luo, Zihao Wan, Ziyue Wang, Yaluo Liu, Pau Tong Lin Xu, Xuanjia Qiao, Xiaolong Wang, Peng Li, Yang Liu. Findings of ACL 2026. [arXiv] [code] [Hugging Face]
Scaffolding Coordinates to Promote Vision-Language Coordination in Large Multi-Modal Models
Xuanyu Lei, Zonghan Yang, Xinrui Chen, Peng Li, Yang Liu. COLING 2025, 2886-2903. [pdf] [arXiv] [code] [Project Webpage]
MLLMs
AdaMMS: Model Merging for Heterogeneous Multimodal Large Language Models with Unsupervised Coefficient Optimization
Model Composition for Multimodal Large Language Models
LVAgent: Long Video Understanding by Multi-Round Dynamical Collaboration of MLLM Agents
VideoChat-M1: Collaborative Policy Planning for Video Understanding via Multi-Agent Reinforcement Learning
Weakly Supervised Vision-and-Language Pre-training with Relative Representations
Datasets
How Do Multimodal Large Language Models Handle Complex Multimodal Reasoning? Placing Them in An Extensible Escape Game
Evaluating Time Awareness and Cross-modal Active Perception of Large Models via 4D Escape Room Task
StreamingBench: Assessing the Gap for MLLMs to Achieve Streaming Video Understanding
Junming Lin, Zheng Fang, Chi Chen, Haoxuan Cheng, Zihao Wan, Fuwen Luo, Ziyue Wang, Peng Li, Yang Liu, Maosong Sun. ICASSP 2026, 12147-12151. [pdf] [arXiv] [code] [Project Webpage]
ActiView: Evaluating Active Perception Ability for Multimodal Large Language Models
CoSpace: Benchmarking Continuous Space Perception Ability for Vision-Language Models
Yiqi Zhu, Ziyue Wang, Can Zhang, Peng Li, Yang Liu. CVPR 2025, 29569-29579. [pdf] [arXiv] [code] [Project Webpage]
CODIS: Benchmarking Context-Dependent Visual Comprehension for Multimodal Large Language Models
Fuwen Luo, Chi Chen, Zihao Wan, Zhaolu Kang, Qidong Yan, Yingjie Li, Xiaolong Wang, Siyu Wang, Ziyue Wang, Xiaoyue Mi, Peng Li, Ning Ma, Maosong Sun, Yang Liu. ACL 2024, 10639-10659. [pdf] [arXiv] [code] [Project Webpage]
StableToolBench: Towards Stable Large-Scale Benchmarking on Tool Learning of Large Language Models
Zhicheng Guo, Sijie Cheng, Hao Wang, Shihao Liang, Yujia Qin, Peng Li, Zhiyuan Liu, Maosong Sun, Yang Liu. Findings of ACL 2024, 11143-11156. [pdf] [arXiv] [code] [Project Webpage]
DocRED: A Large-Scale Document-Level Relation Extraction Dataset
Yuan Yao, Deming Ye, Peng Li, Xu Han, Yankai Lin, Zhenghao Liu, Zhiyuan Liu, Lixin Huang, Jie Zhou, Maosong Sun. ACL 2019, 764-777. [pdf] [code & data] [Leaderboard]
Honors and Awards
- ICAIS 2025 Responsible AI Research Award (2025)
- ACL 2023 Outstanding Paper Award (2023)
- CIKM 2021 Best Resource Paper Nomination (2021)
- First Prize of the Qian Weichang Chinese Information Processing Science and Technology Award (2020)
Professional Services
- Senior Area Chair: AACL (2022, 2026)
- Area Chair: ICLR (2026), ACL (2024–2025), EMNLP (2024), NAACL (2024–2025), EACL (2024, 2026), COLING (2022), AACL (2025)
- Action Editor: ACL Rolling Review (ARR), TMLR
- PC Chair: YWCL (2010)
- Tutorial Chair: CCL (2023), CCMT (2023, 2025)
- Demonstration Chair: CCL (2024)
- Frontier Forum Chair: CCMT (2024)
- Publicity Chair: CCL (2025)
- Student Symposium Chair: CCL (2026)
- Reviewer / PC Member: ICML (2022–2026), ICLR (2024–2025), NeurIPS (2022–2025), ACL (2014, 2021–2023), EMNLP (2014, 2021–2023, 2025), NAACL (2018), CVPR (2023–2026), ICCV (2023, 2025), ECCV (2024, 2026), AAAI (2022–2025), IJCAI (2022–2024), COLING (2020, 2024, 2025), BMVC (2024), COLM (2024–2025), ICRA (2025-2026), IROS (2026), ACMMM (2025-2026), COLM (2024-2025), ACM TIST (2015), TACL (Standing Reviewer), Acta Automatica Sinica, Journal of Computer Science and Technology (JCST)
Education
- Ph.D., Computer Science and Technology, Tsinghua University, Beijing, China (Aug 2009 – Jan 2015)
- B.S., Computer Science and Technology, Tsinghua University, Beijing, China (Aug 2005 – Jul 2009)