Zekun Li 李泽坤
Research Scientist, Google DeepMind
研究科学家，Google DeepMind
I am a Research Scientist at Google DeepMind, working on Gemini post-training and agentic RL
for tool use, planning, and reasoning in LLM agents.
I got my Ph.D. from UC Santa Barbara in 2025, advised by
William Wang and
Xifeng Yan.
My research spans agent capability (post-training, tool use, planning) and
trustworthiness (safety, robustness, multimodal forgery detection).
Previously I was Founding Scientist at Alpha Design AI (ChipAgents),
working on AI-driven chip design.
我是 Google DeepMind 研究科学家，从事 Gemini 后训练与 Agentic RL，
聚焦大模型智能体的工具使用、规划与推理。
2025 年获加州大学圣塔芭芭拉分校计算机科学博士学位，导师为
William Wang 和
Xifeng Yan。
研究覆盖智能体的能力（后训练、工具使用与规划）与可信（安全、鲁棒性、多模态鉴伪）两条主线。
此前担任 Alpha Design AI（ChipAgents）创始科学家，负责 AI 芯片设计。
Highlights 亮点
- [2026] SOPBench and AgentHallu accepted to EMNLP 2026 Main
- [2026] SOPBench 与 AgentHallu 被 EMNLP 2026 Main 接收
- [2026] Papers accepted at ICLR 2026 (MSB, DevOps-Gym), CVPR 2026 (AVFakeBench), and AAAI 2026 (T²Agent Oral, Can Editing LLMs Inject Harm?)
- [2026] 论文被 ICLR 2026（MSB、DevOps-Gym）、CVPR 2026（AVFakeBench）与 AAAI 2026（T²Agent Oral、Can Editing LLMs Inject Harm?）接收
- [2025] Received ACL Outstanding Paper Award (26/8000)
- [2025] 获得 ACL 杰出论文奖（26/8000）
- [2025] Joined Google DeepMind as Research Scientist, working on Gemini
- [2025] 加入 Google DeepMind 担任研究科学家，从事 Gemini 研究
- [2025] Created MassGen (multi-agent scaling system), 1k+ GitHub stars
- [2025] 创建 MassGen（多智能体扩展系统），1k+ GitHub stars
- [2024] Awarded J.P. Morgan AI PhD Fellowship (~10 recipients worldwide per year)
- [2024] 获得摩根大通 AI 博士研究奖学金（每年全球约 10 位获奖者）
Publications 发表论文
h-index: 23 · 2,800+ citations · Google Scholar2026
Zekun Li, Shinda Huang, Jiangtian Wang, Nathan Zhang, Antonis Antoniades, Wenyue Hua, Kaijie Zhu, Sirui Zeng, Chi Wang, William Yang Wang, Xifeng Yan.
"SOPBench: Evaluating Language Agents at Following Standard Operating Procedures and Constraints"
EMNLP 2026 Main
paper | code
paper | code
Xuannan Liu, Xiao Yang, Zekun Li, Peipei Li, Ran He.
"AgentHallu: Benchmarking Automated Hallucination Attribution of LLM-based Agents"
EMNLP 2026 Main
paper | website
paper | website
Tengxiao Liu, Deepak Nathani, Zekun Li, Kevin Yang, William Yang Wang.
"WildSci: Advancing Scientific Reasoning from In-the-Wild Literature"
ACL 2026 Findings
Shuhan Xia, Peipei Li, Xuannan Liu, Dongsen Zhang, Xinyu Guo, Zekun Li.
"AVFakeBench: A Comprehensive Audio-Video Forgery Detection Benchmark for AV-LMMs"
CVPR 2026
Dongsen Zhang, Zekun Li, Xu Luo, Xuannan Liu, Peipei Li, Wenjun Xu.
"MCP Security Bench (MSB): Benchmarking Attacks Against Model Context Protocol in LLM Agents"
ICLR 2026
Yuheng Tang, Kaijie Zhu, ..., Zekun Li, et al.
"DevOps-Gym: Benchmarking AI Agents in Software DevOps Cycle"
ICLR 2026
Xing Cui, Yueying Zou, Zekun Li, et al.
"T²Agent: A Tool-augmented Multimodal Misinformation Detection Agent with MCTS"
AAAI 2026 (Oral)
Canyu Chen*, Baixiang Huang*, Zekun Li, et al.
"Can Editing LLMs Inject Harm?"
AAAI 2026
2025
Xuannan Liu*, Zekun Li*, et al.
"Video-SafetyBench: A Benchmark for Safety Evaluation of Video LVLMs"
NeurIPS 2025 DB
Chenghanyu Zhang*, Zekun Li*, et al.
"SpineBench: Benchmarking Multimodal LLMs for Spinal Pathology Analysis"
ACM Multimedia DB, 2025
Yexiang Liu, Zekun Li, Zhi Fang, Nan Xu, Ran He, Tieniu Tan.
"Rethinking the Role of Prompting Strategies in LLM Test-Time Scaling"
ACL, 2025
ACL Outstanding Paper (26/8000)
Yexiang Liu, Jie Cao, Zekun Li, Ran He, Tieniu Tan.
"Breaking Mental Set to Improve Reasoning through Diverse Multi-Agent Debate"
ICLR 2025
Xuannan Liu, Zekun Li, et al.
"MMFakeBench: A Mixed-Source Multimodal Misinformation Detection Benchmark for LVLMs"
ICLR 2025
Zekun Li, Xianjun Yang, et al.
"MMSci: A Multimodal Multi-Discipline Dataset for PhD-Level Scientific Comprehension"
AI4MAT@ICLR 2025 (Spotlights)
Xinlu Zhang, Chenxin Tian, Xianjun Yang, Lichang Chen, Zekun Li, Linda Ruth Petzold.
"AlpaCare: Instruction-tuned Large Language Models for Medical Application"
SciFM@ICLR 2025
2024
Zekun Li, Baolin Peng, Pengcheng He, Xifeng Yan.
"Evaluating the Instruction-Following Robustness of Large Language Models to Prompt Injection"
EMNLP 2024 Main
Xing Cui, Peipei Li, Zekun Li, Xuannan Liu, Yueying Zou, Zhaofeng He.
"Localize, Understand, Collaborate: Semantic-Aware Dragging via Intention Reasoner"
NeurIPS 2024
Zekun Li, Zhiyu Chen, Mike Ross, et al.
"Large Language Models as Zero-shot Dialogue State Tracker through Function Calling"
ACL 2024 Main
Xuannan Liu, Peipei Li, Huaibo Huang, Zekun Li, et al.
"FKA-Owl: Advancing Multimodal Fake News Detection through Knowledge-Augmented LVLM"
ACM Multimedia 2024
Xing Cui, Zekun Li, Peipei Li, Huaibo Huang, Zhaofeng He.
"InstaStyle: Inversion Noise of a Stylized Image is Secretly a Style Adviser"
ECCV 2024
Shiyang Li, Jianshu Chen, et al., Zekun Li, et al.
"Explanations from Large Language Models Make Small Reasoners Better"
SAI@AAAI 2024
2023
Zekun Li, Baolin Peng, Pengcheng He, Michel Galley, Jianfeng Gao, Xifeng Yan.
"Guiding Large Language Models via Directional Stimulus Prompting"
NeurIPS 2023
Forbes | Wikipedia | Prompting Guide
Forbes | Wikipedia | Prompting Guide
Zekun Li, Shiyang Li, Xifeng Yan.
"Time Series as Images: Vision Transformer for Irregularly Sampled Time Series"
NeurIPS 2023
Xing Cui*, Zekun Li*, Peipei Li, Yibo Hu, Hailin Shi, Zhaofeng He.
"ChatEdit: Towards Multi-turn Interactive Facial Image Editing via Dialogue"
EMNLP 2023 Main
Jing Qian*, Hong Wang*, Zekun Li, Shiyang Li, Xifeng Yan.
"Limitations of Language Models in Arithmetic and Symbolic Induction"
ACL 2023 Main
2022 & Earlier
Zekun Li, Wenhu Chen, Shiyang Li, Hong Wang, Jing Qian, Xifeng Yan.
"Controllable Dialogue Simulation with In-Context Learning"
EMNLP 2022 Findings
Zekun Li*, Hong Wang*, et al.
"Making Something out of Nothing: Building Robust Task-oriented Dialogue Systems from Scratch"
1st Proceedings of Alexa Prize TaskBot (2021)
Zeyu Cui*, Zekun Li*, Shu Wu, Xiaoyu Zhang, Qiang Liu, Liang Wang, Mengmeng Ai.
"DyGCN: Dynamic Graph Embedding with Graph Convolutional Network"
IEEE TNNLS 2022
Xiaoyu Zhang, Haichao Shi, Changsheng Li, Peng Li, Zekun Li, Peng Ren.
"Weakly-supervised Action Localization via Embedding-Modeling Iterative Optimization"
Pattern Recognition 2021
Yujia Zheng, Siyi Liu, Zekun Li, Shu Wu.
"Cold-start Sequential Recommendation via Meta Learner"
AAAI 2021
Yujia Zheng, Siyi Liu, Zekun Li, Shu Wu.
"DGTN: Dual-channel Graph Transition Network for Session-based Recommendation"
NeuRec@ICDM 2020
Zekun Li*, Zeyu Cui*, Shu Wu, Xiaoyu Zhang, Liang Wang.
"Fi-GNN: Modeling Feature Interactions via Graph Neural Networks for CTR Prediction"
CIKM 2019
Zekun Li*, Zeyu Cui*, Shu Wu, Xiaoyu Zhang, Liang Wang.
"Semi-supervised Compatibility Learning across Categories for Clothing Matching"
ICME 2019
Zeyu Cui*, Zekun Li*, Shu Wu, Xiaoyu Zhang, Liang Wang.
"Dressing as a Whole: Outfit Compatibility Learning Based on Node-wise Graph Neural Networks"
WWW 2019
Xuemeng Song, Fuli Feng, Jinhuan Liu, Zekun Li, Liqiang Nie, Jun Ma.
"NeuroStylist: Neural Compatibility Modeling for Clothing Matching"
ACM Multimedia 2017
Open Source 开源项目
MassGen: Multi-Agent Scaling System for GenAI
MassGen：面向 GenAI 的多智能体扩展系统
July 2025 – Present
2025.07 – 至今
Creator and maintainer. A collaborative multi-agent system for complex tasks, supporting multi-model
collaboration, parallel reasoning, and result aggregation. 1k+ GitHub stars.
创建者与维护者。面向复杂任务的协作式多智能体系统，支持多模型协同、并行推理与结果聚合。GitHub 1k+ stars。
Awards 获奖经历
ACL Outstanding Paper Award (26/8000)
ACL 杰出论文奖（26/8000）
ACL, 2025
J.P. Morgan AI PhD Fellowship (~10 recipients worldwide per year)
摩根大通 AI 博士研究奖学金（每年全球约 10 位获奖者）
J.P. Morgan, 2024
1st Prize, Alexa Prize SocialBot Grand Challenge 5
Alexa Prize SocialBot Grand Challenge 5 总冠军
Amazon, 2023
Top-5 Finalist, Alexa Prize TaskBot Challenge
Alexa Prize TaskBot Challenge 决赛前 5 名
Amazon, 2022
Academic Excellence Fellowship
学术卓越奖学金
UCSB, 2021