Description

A live reinforcement learning environment where LLMs play a 10-player social deduction game against hundreds of thousands of human players. The system uses a multi-layer harness (strategy, planning, and chat layers) with a custom grammar to convert model responses into game-executable actions and evaluates models based on win-rate impact and behavioral signals.

Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.