Federal Grant Award Summary

The National Science Foundation (NSF) Division of Computing and Communication Foundations awarded a $600,000 Project Grant to the University of Washington on September 15, 2026, under the Computer and Information Science and Engineering program (CFDA 47.070). This collaborative research project, scheduled for completion by August 31, 2029, develops artificial intelligence methods to reduce hallucinations and improve the reliability and transparency of Large Language Models (LLMs) in health information applications. The project delivers three primary products: (1) a multilingual evidence-aligned knowledge repository with reliability scoring across English, Spanish, and French, supplemented by dynamic benchmarks testing reasoning, attribution, abstention, and clarification capabilities; (2) new model training and inference methods enabling non-hallucinating long-form generation, fine-grained attribution, calibrated uncertainty estimation, and proactive clarification questioning; and (3) enhanced LLM behavior for high-stakes environments.

The project focuses on women's health conditions—including breast cancer, osteoporosis, cardiovascular disease, autoimmune disorders, and mental health—as a testbed for advancing LLM reliability across diverse medical domains. By addressing the critical problem of LLMs expressing erroneous information with high confidence, this research aims to accelerate the adoption of artificial intelligence technologies in high-risk sectors including medicine and law, where stable and transparent model behavior is essential for public safety and decision-making.