REU site: knowledge beyond language with vector embeddings

The National Science Foundation Division of Computing and Communication Foundations awarded the University of North Texas Research and Innovation Division $464,978 on October 1, 2026, for a three-year Research Experiences for Undergraduates (REU) site focused on knowledge transfer across AI domains through vector embeddings.

The program will support ten undergraduate students annually for ten weeks each summer through September 30, 2029. Students will train in building AI systems capable of sharing knowledge across domains by translating information—images, text, or sound—into numerical vector representations that preserve nuanced understanding learned by deep learning models and make it accessible for other applications. Each student will identify a project domain and work with a faculty advisor on research that creates vector representations of entities appearing across multiple domains, applies these embeddings to improve prediction models, and contributes the work to a shared, reusable knowledge base. Students will participate in an existing summer AI research program integrating current university students and external REU participants to facilitate cross-departmental collaboration.

The award is funded under Computer and Information Science and Engineering (CFDA 47.070), which supports investigator-initiated research and education across computing, communications, and information science and engineering. Place of performance is Denton, Texas.

Abstract

This three-year research experiences for undergraduates (REU) site at the university of north texas will support 10 students for 10 weeks each summer and train them to build artificial intelligence (AI) systems capable of sharing knowledge across domains through vector embeddings. Vector embeddings translate information (e.g., images, text or sound) into numerical data. These data are then converted into lists that allow the numbers to convey context and meaning. Current AI systems who are working in areas such as visual object recognition, speech recognition, and understanding natural language require extensive training, with much of the learned knowledge locked within the structure of the system, which is then difficult to reuse. However, students in this program will focus on creating and leveraging highly-trained AI systems to represent information in ways that preserve the nuanced understanding learned by these systems and make IT accessible for other applications. This REU brings together an interdisciplinary team to support projects that showcase the benefits of AI systems that can exchange and reuse learned knowledge. Early in the program, each student will identify a project domain and a faculty advisor with whom they can work. Students will also participate in a long-standing AI summer research program integrating current university students and external REU students to facilitate collaboration across departments and student expertise. Specifically, the training in this REU will allow students to more efficiently represent and transfer the knowledge acquired by self-supervised deep learning models. Each year, students will create vector representations of entities that appear across multiple domains, apply these embeddings to improve prediction models, and systematically evaluate, document, and contribute them to a shared, reusable knowledge base. These efforts are coordinated through common documentation, evaluation, and sharing practices that enable comparison and reuse of embeddings across projects. For the first five weeks, the students will be exposed to different embedding strategies and machine learning applications that use them, then transition to developing, testing, and refining their individual research efforts in the last five weeks. This REU will help prepare a workforce of students not only adept at using deep learning models but also capable of extending their functionality through reusable and shareable representations. Additionally, this project will train a diverse range of students from college partners with limited research resources to work in interdisciplinary teams at a carnegie R1 research institution. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.- subawards are not planned for this award.