ResearchBenchmarksMulti-Agent Strategic Reasoning on Tic-Tac-Toe (In-domain)Multi-Agent Strategic Reasoning on Tic-Tac-Toe (In-domain)Follow67.2Success RateMAGE-1.02416.68834.452.112Mar 4, 2026Evaluation resultsMethodMethodLinksSuccess RateMAGECategory=Meta-RLCategory=Meta-RL2026.0367.2LAMERCategory=Meta-RLCategory=Meta-RL2026.0360.2GiGPOCategory=RLCategory=RL2026.0341.4ReflexionCategory=In-Context Le...Category=In-Context Learning2026.0324.2GRPOCategory=RLCategory=RL2026.0321.9ReActCategory=In-Context Le...Category=In-Context Learning2026.033.9MementoCategory=Memory-basedCategory=Memory-based2026.033.1A-MEMCategory=Memory-basedCategory=Memory-based2026.031.6