FutureSim: Replaying World Events to Evaluate Adaptive Agents

TL;DR

FutureSim is a simulation framework that replays real-world events to evaluate AI agents' ability to predict and adapt to unfolding world developments over extended periods.

Contribution

The paper introduces FutureSim, a novel benchmark for assessing AI agents' long-term prediction and adaptation in realistic, dynamic environments using chronological event replay.

Findings

Best agent achieved 25% accuracy in predicting future events.

Many agents performed worse than no prediction, highlighting challenges.

FutureSim enables studying long-horizon adaptation, search, memory, and uncertainty reasoning.

Abstract

AI agents are being increasingly deployed in dynamic, open-ended environments that require adapting to new information as it arrives. To efficiently measure this capability for realistic use-cases, we propose building grounded simulations that replay real-world events in the order they occurred. We build FutureSim, where agents forecast world events beyond their knowledge cutoff while interacting with a chronological replay of the world: real news articles arriving and questions resolving over the simulated period. We evaluate frontier agents in their native harness, testing their ability to predict world events over a three-month period from January to March 2026. FutureSim reveals a clear separation in their capabilities, with the best agent's accuracy being 25%, and many having worse Brier skill score than making no prediction at all. Through careful ablations, we show how FutureSim offers a realistic setting to study emerging research directions like long-horizon test-time adaptation, search, memory, and reasoning about uncertainty. Overall, we hope our benchmark design paves the way to measure AI progress on open-ended adaptation spanning long time-horizons in the real world.