LLM agents play Pokemon showdown

A multi-agent environment was built on Pokémon Showdown's 2v2 format to evaluate LLM agents' collaborative abilities, focusing on communication and negotiation. This setup allows two Qwen3.8-27B agents to act as teammates, making strategic decisions like splitting targets and correcting each other's type-matchup mistakes, as seen in a Turn 1 battle excerpt. The goal is to measure and later train these collaborative behaviors in a game-theory environment where opponents' sets are hidden and damage is random.