AI TrackerFrontier desk · AEST
Back to news
normalaitech

A Unified Arena for LLM Agents across Diverse Eurogame

arXiv cs.AI · 27 Oct 2026, 03:14 AEST

A new study introduces DECE, a unified arena for evaluating large language model agents across ten Eurogames. The research highlights the importance of competitive performance and resource efficiency in assessing agent configurations.

Key facts

Read aloud

A new study introduces DECE, a unified arena for evaluating large language model agents across ten Eurogames. The research highlights the importance of competitive performance and resource efficiency in assessing agent configurations. Interactive environments offer a practical way to evaluate large language model L L M agents, and competitive games extend this evaluation to decisions shaped by competing goals and other agents' choices. We introduce DECE Diverse Eurogame Competitive Environments, a unified arena spanning ten Eurogames with shared interfaces and evaluation protocols. Our primary study evaluates twelve model-harness configurations, spanning nine models and four execution harnesses, in four hundred eighty three-player matches across ten games. Matched studies show that the same model can perform differently across execution harnesses, adding another source of variation to these profiles. Together, the findings support evaluating complete agent configurations by both their competitive performance across games and the resources needed to achieve it.

Sources in this cluster