NSF award 2619180 · Georgia Tech Research Corporation

SHF: Formal Reasoning of Reinforcement Learning Policies

Final report due

29 December 2030

in 1,533 days · The award ends on 31 August 2030.

GrantClerk is a private service. We are not NSF or any government agency.

Dates this award owes

NSF emails you about this too.

Decide about an extension3 May 2030in 1293 days
award ending 2030-08-31 · Research.gov
This one is ours, not NSF’s.
Extension request17 Jul 2030in 1368 days
award ending 2030-08-31 · Research.gov
PAPPG 24-1: “at least 45 days prior to the end date”
No-cost extension notice21 Aug 2030in 1403 days
award ending 2030-08-31 · Research.gov
PAPPG 24-1: “at least ten calendar days prior to the end date”
Final report due29 Dec 2030in 1533 days
award ending 2030-08-31 · Research.gov
PAPPG 24-1: “no later than 120-days following the end date of the award”
Project outcomes report due29 Dec 2030in 1533 days
award ending 2030-08-31 · Research.gov
PAPPG 24-1: “no later than 120 days following the end date”
Award is closed out29 Dec 2030in 1533 days
award ending 2030-08-31 · Research.gov
PAPPG 24-1: “financially closed out 120 days after their expiration date”

Annual report dates come from the budget period on your award notice. That date is not public, so it is not shown here.

How we got this date

NSF lists this award's end date as 31 August 2030.
NSF rules say the final report is due no later than 120 days after that.
So the final report is due by 29 December 2030.

Read the rule on Research.gov

Get a note before each report date

Add this award to your desk. We email you before each date and keep the dates in one place.

Renews each year until you cancel. Cancel any time from your account.

Already a member? Sign in

Award card

Award number: 2619180
Period of support: 1 Sep 2026 – 31 Aug 2030
Amount: $600,000
Awardee: Georgia Tech Research Corporation, Atlanta, GA
Program: Software & Hardware Foundation
Directorate: CSE

Abstract

Artificial intelligence (AI) systems trained through reinforcement learning (RL) are increasingly deployed in safety-critical settings such as self-driving cars, robotic surgery, and autonomous trading, yet these systems operate without any formal guarantees that they will behave correctly. A core problem is that RL systems learn from numerical rewards that can be gamed by the agent in unintended ways — a phenomenon…

Is this right?

We copy this from the public NSF award record. Your award notice and NSF always win. If a date looks wrong, check Research.gov.