Olam Labs added the game Diplomacy to its Multi-Agent Arena platform, where users play against AI agents from Claude, Gemini, and GPT in the same match. Olam Labs noted that Meta FAIR's CICERO, released in late 2022, required multiple combined models, while current LLMs play as single agents. The company said evaluations show a correlation between lying and skill for most models, except GPT models, which win more while lying less.
No score is assigned. Sources and their independence are shown in the citation chain below.