AI Influence Benchmarking & Evaluation

The AI Influence Benchmarking and Evaluation subcontract, issued by the Defense Advanced Research Projects Agency under the U.S. Department of Defense, focuses on core research and development to establish benchmarks for AI systems. The contractor is responsible for developing evaluation frameworks for large language models, designing influence tests to analyze the impact of prompts and training data on outputs, and performing technical feasibility studies regarding benchmark scalability. These activities require the use of high-compute GPU clusters and strict adherence to the NIST AI Risk Management Framework to deliver prototype evaluation metrics and technical reports. This opportunity is a Total Small Business Set-Aside under the SBIR/STTR mandate, restricted to firms with fewer than 500 employees, falling under NAICS code 541715. The solicitation was posted on September 23, 2026, with a response deadline of October 21, 2026. The project supports prime contractors on Department of War SBIR projects to ensure the rigorous testing and validation of artificial intelligence systems.