Federal Project Grant Award Summary

North Carolina State University received a $300,000 Project Grant from the National Science Foundation's Division of Computing and Communication Foundations (Computer and Information Science and Engineering program, CFDA 47.070) effective September 1, 2026, through August 31, 2029. This collaborative research project develops a knowledge augmentation framework to improve model editing in foundation models, including Large Language Models (LLMs). The primary deliverables include methods for both training-time and inference-time editing that enhance the generalizability and reliability of knowledge updates across foundation models. Training-time solutions incorporate visual generation, text augmentation, and graph-guided techniques to identify dependent facts, while inference-time methods employ contrastive decoding, preference-based optimization, and multi-step reasoning models to ensure faithful application of edited facts.

The project will evaluate proposed methods across multiple domains including text-to-image generation, question answering, classification, and completion tasks, with specific applications in medical image diagnosis and geographic remote sensing. Educational activities include training undergraduate and graduate students in reliable foundation models, multimodal learning, and responsible machine learning, with targeted initiatives to broaden participation in computing. The research outcomes are intended to advance understanding of how foundation models store, update, and utilize knowledge while supporting safer and more trustworthy deployment of these systems in critical domains such as healthcare, science, and geospatial analysis.