LLM agents play Pokemon showdown
A source-linked abstract prepared once when this article was collected.
A project tests LLM teammates in a 2v2 Pokémon Showdown environment. Two Qwen3.8-27B teammates with free talk won 64% of battles against a max-damage bot, versus 42% with no talk, 52% with one message each, and 54% when one model controlled both Pokémon. Free talk used about 978k prompt tokens per battle versus 259k with no talk. The same model won 96% against a 2B-model team and 74% against a 9B-model team, which the author presents as evidence the environment can distinguish models. The author plans to separate the effect of talking from seeing a partner's move and to build a leaderboard and RL training.
Why it matters
It offers a controlled way to measure whether LLM agents coordinate under hidden information, a setting relevant to multi-agent systems.
Read with care
Results come from a single model, one scripted opponent, and seeded battles; source claims not independently checked.
Discussion
Comments are reviewed before publication. Discuss the evidence, not the person.
No published comments yet.
Reader discussion opens with accounts. You can read the original source and published comments now.