Rui Hou is an artificial intelligence researcher known for his work in large language models (LLMs), multi-modal representation, and machine learning infrastructure. He is a research scientist at Meta, where he is a member of the company's Generative AI and Meta Superintelligence team. ^{[1] [2]}

Education

Hou received a Bachelor's Degree from Tongji University in China. He later attended the University of Michigan, Ann Arbor, from 2014 to 2020. During his time at the university, he earned a Master's degree in Computer Science as well as a Doctor of Philosophy (Ph.D.) in Intelligent Systems. His doctoral research was conducted under the supervision of advisor Jerome P. Lynch. ^{[3] [4] [2] [7]}

Career

Rui Hou began his professional career with an internship in Building Information Modeling at Shanghai Construction Group Co., Ltd. in 2014. While pursuing his graduate studies at the University of Michigan, he served as a Graduate Student Research Assistant (GSRA) and a Graduate Student Instructor (GSI) from 2014 to 2019. During this period, he gained further industry experience through multiple internships at the Toyota Research Institute, first as a Machine Learning Intern in 2017 and later as a Research Intern in 2019. He also completed an internship as a PhD Software Engineer in machine learning at Facebook in 2019.

After completing his Ph.D. in 2020, Hou joined Meta (formerly Facebook) as a full-time Research Scientist. He advanced within the company, taking on roles as a Senior Staff Research Scientist and Senior Research Scientist Manager. His work at Meta included contributions to the Llama family of large language models, specifically in post-training research. In early 2025, Hou briefly joined xAI as a Member of Technical Staff. He subsequently returned to Meta to become part of its newly established Meta Superintelligence research group.

Hou's research is centered on key areas of artificial intelligence, with a particular focus on large language models and multi-modal representation. His work addresses the challenges of scaling foundation models, enhancing the multilingual capabilities of AI systems, and improving model robustness. He has contributed to research on continual learning for language models, adversarial training techniques, and methods for overcoming vocabulary limitations in multilingual models.

He has co-authored numerous papers presented at major AI conferences. Selected publications include:

• "Layer Swapping for Zero-Shot Cross-Lingual Transfer in Large Language Models" (ICLR, 2025)
• "Law of the Weakest Link: Cross Capabilities of Large Language Models" (ICLR, 2025)
• "Effective Long-Context Scaling of Foundation Models" (NAACL, 2024)
• "XLM-V: Overcoming the Vocabulary Bottleneck in Multilingual Masked Language Models" (EMNLP, 2023)
• "RoAST: Robustifying Language Models via Adversarial Perturbation with Selective Training" (EMNLP, 2023)
• "Progressive Prompts: Continual Learning for Language Models" (ICLR, 2023)
• "Semantically-Guided Representation Learning for Self-Supervised Monocular Depth" (ICLR, 2020)

The above is a partial list of his published works. ^{[3] [2] [4] [5] [7] [8]}

Meta Superintelligence Team

In mid-2025, Hou was named as a member of Meta's newly formed "Superintelligence" team, a strategic initiative aimed at advancing foundational research toward artificial general intelligence (AGI). The group was assembled with researchers and engineers recruited from other leading AI organizations such as OpenAI, DeepMind, and xAI. Hou returned to Meta to join this team as a Research Scientist after his time at xAI. The team's composition, which includes a significant number of research scientists with doctoral degrees, underscores Meta's strategic focus on fundamental research in core AI domains, including LLMs, computer vision, multimodal systems, and AI alignment. ^{[2] [6] [7]}