FM-Bench: Long-Horizon Agent Evaluation via Football Management
August 18, 2026
FM-Bench evaluates agent decision-making over long horizons by simulating 20 years of football club management using 26 tools. The benchmark uses a deterministic engine to track cumulative consequences without relying on human or LLM judges.
HOW THIS AFFECTS YOU
●
builderYou can test how your agents handle cumulative consequences and long-term strategy.
●
researcherThis provides a scalable, objective way to measure long-horizon agentic capabilities.