Dr. Keqian Li is a researcher at the Shanghai Institute of Intelligent Education, focusing on large-scale artificial intelligence systems, multimodal foundation models, and data-driven intelligent learning technologies. He received his Ph.D. in Data Mining and Deep Learning from the University of California, Santa Barbara, and his B.S. in Computer Science from Tsinghua University through the prestigious Yao Class. Prior to joining ECNU, he served as Organization Lead at XPENG, where he led the global multimodal foundation model initiative, building foundational multimodal models from scratch, establishing a unified organization across multimodal intelligence, evaluation, and reinforcement learning, and assembling a world-class team spanning large language models, vision transformers, data, and embodied deployment. From 2022 to 2025, he was a Research Lead at Meta, co-founding the company’s data research organization (now part of FAIR), leading hallucination mitigation efforts for generative AI highlighted in Meta’s earnings reports, and overseeing unified user understanding systems for global monetization and personalization. Earlier, he was a Researcher at Yahoo Research Labs (affiliated with Verizon), leading large-scale data science initiatives across global users.
个人资料
教育经历UC Santa Barbara Doctor of Philosophy (PhD), Data mining, deep learningDoctor of Philosophy (PhD), Data mining, deep learning Tsinghua University Bachelor of Science (B.S.), Pilot Class of Computer Science(Yao Class) 工作经历
XPENG · Full-time Sep 2025 - Dec 2025 · 4 mos Leading the global multimodal foundation model effort at XPeng - Building the original version of XPeng foundational multimoal that match or surpass SOTA pretrained from scratch - Creating the unified organization of across multimodal intelligence, machine intelligence and evaluation and RL - Building a team of world class experts for across full spectrum of foundational modeling. LLM, VIT, pre-training, data, post-training, RL, and deployment on embodied scenarios Research Lead Meta Aug 2022 - Jul 2025 · 3 yrs 4 mos - co-founded the Meta's data research org (now part of Meta's FAIR Lab) - founded the hallucination mitigation for Meta's generative AI highlighted in CFO's earning report - In charge of the unified user understanding system for global monetization and personalization across Meta's asset of media Researcher Yahoo Research Labs, affiliated with Verizon · Full-time Nov 2019 - Jul 2022 · 2 yrs 9 mos Manhattan, New York, United States / Sunnyvale, CaliforniaManhattan, New York, United States / Sunnyvale, California Leading data science for Verizon ecosystem across global users 个人简介社会兼职研究方向
招生与培养开授课程科研项目- 启创 InnoSpark 大模型 学术成果The Llama 4 herd: The beginning of a new era of natively multimodal AI Keqian Li and other contributors Presented at CVPR (Keynote) May 2025 The Llama 3 Herd of Models Keqian Li and other contributors Technical Report. July 2024 CALM: Common-Sense Knowledge Augmentation June 2022 for Document Image Understanding Qinyi Du, Qingqing Wang, Keqian Li, Jidong Tian, Liqiang Xiao, Yaohui Jin. In Proceedings of the 30th ACM International Conference on Multimedia. ( ACM MM 22 • Merchandise ads Collaborated with the Big-Table, GFS , Map-Reduce and Borg team and built the first generation of feature store and ads recommender system for matching google merchandise ads with the users integrating Search, Google Gmail, Google Plus and product knowledge base for user feedback prediction and multichannel ads serving. ). MGEL: Multi-Grained Representation Analysis and Ensemble Learning Jan 2022 for Text Moderation Fei Tan, Changwei Hu, Yifan Hu, Kevin Yen, Zhi Wei, Aasish Pappu, Serim Park, Keqian Li. IEEE Transactions on Neural Networks and Learning Systems ( IEEE TNNLS ). TNT: Text Normalization based Pre-training of Transformers Oct 2020 for Content Moderation Fei Tan, Yifan Hu, Changwei Hu, Keqian Li, Kevin Yen.The 2020 Conference on Empirical Methods in Natural Language Processing. ( EMNLP 20’). HierCon: Hierarchical Organization of Technical Documents Nov 2019 Based on Concepts Keqian Li, Shiyang Li, Semih Yavuz, Hanwen Zha, Yu Su and Xifeng Yan (Best Paper Runner-ups of ICDM 19’ ). The 19th IEEE International Conference on Data Mining (ICDM 19’). Mining Algorithm Roadmap in Scientific Publications Aug 2019 Hanwen Zha, Wenhu Chen, Keqian Li and Xifeng Yan. 25th ACM SIGKDD international conference on Knowledge discovery and data mining (KDD 2019 ). Concept Mining via Embedding Keqian Li, Hanwen Zha, Yu Su, Xifeng Yan. The 18th IEEE International Conference on Data Mining (ICDM 18 ). Nov 2018 PoQaa: Text Mining and Knowledge Sharing for Scientific Publications Aug 2018 Keqian Li, Ping Zhang, Honglei Liu, Hanwen Zha, Xifeng Yan. The 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Min- ing (KDD 18 ) (system). FTS: Faceted Taxonomy Construction and Search for Scientific Publications Aug 2018 Hanwen Zha, Jiaming Shen, Keqian Li, Warren Greiff, Michelle Vanni, Jiawei Han and Xifeng Yan. 24th ACM SIGKDD international conference on Knowledge discovery and data mining (KDD 18 ) (system). Unsupervised Neural Categorization for Scientific Publication Keqian Li , Hanwen Zha, Yu Su, Xifeng Yan. The 18th SIAM International Conference on Data Mining (SDM 18 ). May 2018 Discovering Enterprise Concepts using Spreadsheet Tables Aug 2017 Keqian Li, Yeye He, Kris Ganjam. The 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Min- ing (KDD 17 ). Show Me the Money:Dynamic Revenue-Maximizing Recommendations Wei Lu, Shanshan Chen, Keqian Li, Laks Lakshmanan. Proceedings of the VLDB Endowment 7.14 (VLDB 15 ). July 2015 On Social Event Organization Keqian Li, Wei Lu, Smriti Bhagat, Laks VS Lakshmanan, and Cong Yu. The 20rd ACM SIGKDD International Conference on Knowledge Discovery and Data Min- ing (KDD 14 ). Aug 2014 荣誉及奖励Techpulse Best Internal Talk Award 2022 Best of ICDM at The 19th IEEE International Conference on Data Mining 2019 University of California Regents Fellowship 2016 Tsinghua Academy Awards 2013 |