Mentiss

Mentiss AI is pioneering the future of AI benchmarking through strategic social deduction games. We use Werewolf (Mafia) as a multi-agent, zero-sum environment to evaluate AI models' reasoning, persuasion, and social intelligence capabilities. Unlike traditional benchmarks that rely on self-reported improvements, our platform provides objective, measurable outcomes through: • Large-scale game simulations (tens of thousands of matches) • Role-specific performance tracking across different AI models • Step-level decision accuracy measurement • Social intelligence metrics including trust, logic consistency, and emotional alignment Our platform generates high-quality synthetic training data from gameplay logs, enabling AI models to learn from strategic decision-making scenarios. We're building the bridge from language intelligence to social intelligence. Perfect for AI companies seeking comprehensive model evaluation before product launch.#AI #MachineLearning #Benchmarking #SocialIntelligence #Werewolf #Mafia #MultiAgent #LLM