Agents

An experimental analysis of coordination capabilities in large language models through a simple pattern discovery game, revealing that same-model pairs achieve significantly better coordination than cross-model pairs.