Project Grant N000142412704

This Project Grant award of $300,000 from the National Science Foundation's (NSF) Computer and Information Science and Engineering (CISE) program supports research at Cornell University to develop improved evaluation concepts and technologies for large language models (LLMs). The project aims to discover a taxonomy of "evaluation concepts" that can assess LLM responses for criteria like factuality, level of detail, and alignment with user needs. It will then develop customized reward...

This National Science Foundation (NSF) Computer and Information Science and Engineering (CISE) Federal Grant Program (CFDA 47.070) award of $799,999 to the University of Washington provides funding for a 4-year collaborative research project focused on developing tools and methods to leverage large language models (LLMs) in support of divergent, convergent, and cooperative work. The key objectives of the project are to: 1) Develop design guidance and workflows for building more reliable and...

This Project Grant award, valued at $158,664 and awarded on June 15, 2025, is funded by the National Science Foundation's (NSF) Integrative Activities program (CFDA 47.083). The primary objective is to develop methods for using large language models (LMs) to apply existing written guidelines, such as grading rubrics and job listings, to assist with decision-making tasks like essay grading and resume screening. The project will first create datasets and annotation tools to capture how humans link...

This $300,000 Project Grant award from the National Science Foundation's (NSF) Computer and Information Science and Engineering (CISE) program supports research to improve the evaluation and functionality of large language models (LLMs). The University of Texas at Austin will lead this 3-year collaborative research project to identify a taxonomy of "evaluation concepts" for assessing LLM responses and develop technology to automatically evaluate and improve LLM performance based on...

The National Science Foundation (NSF) awarded a $600,000 Project Grant under the Computer and Information Science and Engineering (CISE) program to The Regents of the University of California, doing business as the University of California, Berkeley. This 3-year award aims to integrate semantic knowledge from large language models, such as ChatGPT, into automated decision-making and control systems to leverage their predictive capabilities while maintaining reliability and robustness. The...

The National Science Foundation has awarded a $299,990 Project Grant to Columbia University to develop a prototype system using large language models to assist in the review process for releasing government records. The project aims to determine the additional data and training required to achieve acceptable levels of accuracy in using large language models to identify information that is already in the public domain, in order to streamline the document review process. The iterative system...

This Project Grant award of $360,000.00 from the National Science Foundation (NSF) Computer and Information Science and Engineering (CISE) program (CFDA 47.070) to the Massachusetts Institute of Technology (MIT) aims to develop new methods for efficient, architecture-aware algorithms for large language models (LLMs). The goals are to make existing LLM applications more efficient, enable new applications, and broaden access to this transformative AI technology. The key focus areas include: (1)...

This three-year $800,000 project grant from the National Science Foundation's Computer and Information Science and Engineering program aims to advance understanding of large language models through mathematical and conceptual analysis. The Trustees of Princeton University will receive funding to develop simplified generative text models, analyze how language models are trained on such generated texts, examine why learned models can perform downstream tasks, and design new adaptation methods with...

This Project Grant award of $300,000 from the National Science Foundation (NSF) Technology, Innovation, and Partnerships (CFDA 47.084) program supports a study at Georgetown University to develop an evaluation methodology for measuring the impacts of implementing large language model (LLM)-based tools to assist human experts working in federal, state, and local government programs. The project will compare the performance of LLM-only, human-only, and human-LLM hybrid responses across key metrics...

This Project Grant award of $500,000 from the National Science Foundation's (NSF) Computer and Information Science and Engineering (CISE) program aims to leverage the FABRIC national research infrastructure to efficiently and securely train language models and perform inference. The key objectives are to: (a) develop techniques for efficient language model training on heterogeneous GPU clusters and high-speed networks, (b) implement scalable and secure processing of training jobs using...