Federal Project Grant Award Summary

The National Science Foundation's Division of Computing and Communication Foundations awarded $417,565 to Washington University (awarded October 1, 2026; completion September 30, 2031) under the Computer and Information Science and Engineering program (CFDA 47.070) to develop an adaptive knowledge synthesis framework for large language models (LLMs). The project addresses critical gaps in AI systems' ability to tackle complex, knowledge-intensive tasks requiring multi-source information gathering and synthesis. The research will establish a proactive, efficient, and transparent reasoning framework that enables LLMs to formulate structured research plans through hypothesis-driven decomposition, with a conductor model orchestrating adaptive expertise across different problem-solving phases. This technology directly supports evidence-based fields such as legal analysis and scientific discovery by creating more reliable and transparent decision-making tools.

In addition to the core research activities organized across three integrated technical thrusts, the project includes substantial educational and outreach components. Washington University will develop new university courses teaching critical analysis of AI systems and lead hands-on outreach activities for local school students to inspire future researchers in the field. The comprehensive approach combines technological advancement in AI reasoning capabilities with workforce development initiatives to broaden participation in computing research.