arXivAustin Tudor David Andrews, Liam Wilkinson, Jamie Heagerty, Harry Coppock, Jakob Nicolaus Foerster, Rui Ponte CostaWed, Sep 2, 2026, 4:31 AM PDT
score 16.3
New benchmark tests AI agents' long-term planning using strategy game Civilization VI
Original: CivBench: A Long-Horizon Benchmark for Tool-Mediated Agents in Civilization VI
Source: arxiv.org ↗
Writing ELI5 summary…