Exponentiating Mathematics (expMath)

The Defense Advanced Research Projects Agency (DARPA), through its Information Innovation Office (I2O), has announced the Exponentiating Mathematics (expMath) program under Broad Agency Announcement HR001125S0010. This funding opportunity seeks innovative research in artificial intelligence (AI) to significantly accelerate progress in mathematics. DARPA, known for pioneering technologies in national security, envisions transforming mathematical practices to enhance national competitiveness and address complex scientific challenges. The core objective of the expMath program is to develop an AI collaborator capable of increasing the rate of progress in mathematics by orders of magnitude. The initiative emphasizes advancing AI's ability to handle abstraction, particularly in formulating and proving lemmas, which are crucial in mathematical reasoning. The program includes two technical areas (TAs): TA1 focuses on automating the decomposition of mathematical problems and the formalization/informalization of proofs, while TA2 centers on evaluating AI performance against professional-level mathematics benchmarks. Proposers may apply to one or both TAs but must submit separate proposals for each. Eligible applicants include a broad range of entities, both U.S. and non-U.S., provided they comply with applicable laws and agreements. This includes universities, businesses, and nonprofits, while certain limitations apply to UARCs and FFRDCs. Proposals should not be classified, and the work is expected to remain unclassified. Submissions must include detailed technical approaches, cost structures, and supporting documentation. An abstract phase is available, though optional, and uses required templates. All submissions must be made via the Broad Agency Announcement Tool (BAAT) by the designated deadlines. Evaluation criteria for proposals include scientific and technical merit, relevance to DARPA’s mission, and cost realism. DARPA encourages open-source development and the use of liberal licenses (e.g., Apache, MIT), and collaboration among teams is facilitated by Associate Contractor Agreements. Successful proposals are expected to follow a 36-month schedule, starting January 1, 2026, with semiannual evaluations and mandatory participation in PI meetings and site visits. Important dates include the proposers day on April 23, 2025, abstract submission deadline on May 15, 2025, and the final proposal submission deadline extended to July 15, 2025. Questions must be submitted by June 25, 2025. The anticipated award instruments include procurement contracts, cooperative agreements, and other transactions, and multiple awards are expected. For technical and administrative questions, contact expMath@darpa.mil. Additional resources, including proposal templates and FAQs, are available on the expMath program page on DARPA’s website.

Submitting an abstract is strongly encouraged for early feedback Use DARPAConnect for free resources and guidance Follow all formatting and submission guidelines strictly via BAAT