Game Arena拓展AI基准测试，新增扑克与狼人杀
📄 中文摘要
Game Arena平台正通过引入扑克和狼人杀两款新游戏，显著扩展其人工智能基准测试能力。此举旨在为评估AI模型在更复杂、更具战略性和社会交互性的环境中的表现提供更丰富的场景。扑克游戏，特别是无限注德州扑克，需要AI具备强大的博弈论推理、不完全信息下的决策能力、风险评估以及对手建模能力。AI必须能够理解并利用概率、预测对手行为、管理资金，并在诈唬和反诈唬中展现出高超的策略。而狼人杀则进一步挑战了AI在自然语言理解、社会推理、谎言识别、说服与协商等方面的能力。AI玩家需要分析对话、识别矛盾、建立信任关系或制造不信任，并在复杂的社会动态中形成联盟或揭露身份。
📄 English Summary
Advancing AI benchmarking with Game Arena
Game Arena is significantly expanding its AI benchmarking capabilities by introducing two new games: Poker and Werewolf. This initiative aims to provide richer scenarios for evaluating AI models' performance in more complex, strategic, and socially interactive environments. Poker, particularly No-Limit Texas Hold'em, demands AI to possess strong game theory reasoning, decision-making under incomplete information, risk assessment, and opponent modeling capabilities. AI must be able to understand and leverage probabilities, predict opponent behavior, manage finances, and demonstrate sophisticated strategies in bluffing and counter-bluffing. Werewolf further challenges AI in natural language understanding, social reasoning, deception detection, persuasion, and negotiation. AI players need to analyze conversations, identify inconsistencies, build trust or distrust, and form alliances or uncover identities within complex social dynamics. The inclusion of these new games will foster advancements in cutting-edge areas such as multi-agent systems, imperfect information games, and human behavior simulation. Concurrently, AI models Gemini 3 Pro and Flash have topped Game Arena's chess leaderboard, indicating their exceptional decision-making and planning abilities in traditional strategic games. Game Arena's continuous expansion offers researchers and developers a dynamic testing platform to measure and enhance AI's intelligence levels in increasingly complex and real-world scenarios, driving breakthroughs in broader applications of AI technology.