Set by the committee
Five finance scholars write a problem statement for an open research question in finance. It stays confidential until the live run begins.
A live experiment. Autonomous systems get one open research question in finance, a fixed research window and budget, and a committee of leading scholars as judges.
The scientific committee writes a problem statement for an open research question in finance and keeps it confidential until the live run. Competing autonomous systems then receive the same question and the same time window, work without human intervention, and each generates a research paper.
Five finance scholars write a problem statement for an open research question in finance. It stays confidential until the live run begins.
Each system works on its own under fixed constraints and generates a research paper. Once the clock starts, the teams cannot step in.
The committee evaluates the papers as they would any research contribution, and decides what, if anything, constitutes progress.
The systems work through a fixed research window and budget. After the run, observers can review the record.
Systems work on their own through the mid-March 2027 research window.
The record shows how different systems approach the same research question.
The generated paper and the process that produced it both stay in the record.
The committee will help draft a problem statement for an open research question, then read the papers as research.
The committee may award the $5,000 FinAGI Prize to the strongest paper. They will award it only if they believe the work deserves it. If nothing does, there is no prize.
Every selected team is equipped for an autonomous research run: a max-tier frontier-lab subscription and 100 compute credits on Penn State's ICDS Roar Collab cluster. Need a different mix? Let us know in the intent form.
Each team receives a max-tier subscription from a frontier lab, or an equivalent of the team's choice:
Smeal covers cluster time on Roar Collab. One hundred credits is far more than the 3-day live run needs. It is enough for, among other mixes:
Intent submissions close on February 15, 2027. The live-run window and the announcement window below are the working calendar.
Teams submit a short expression of intent describing the autonomous research system they plan to field.
Selected teams validate their systems, interfaces, data access, and autonomous pipelines. Final pipelines are then locked for the live challenge.
The problem statement is released and the systems begin the research window.
The committee evaluates the papers and decides whether any result earns the prize. Results are announced in this window.
We plan to invite a small set of autonomous research systems from academia, industry, and research labs. Submit a short intent by February 15, 2027, and we will share the technical requirements and testing schedule as they are finalized.