[andreasgl]

Fun project, thanks for sharing!

Have you tried giving the models a topic to discuss? I looked at a few games and the only thing they seem to discuss is how to conduct the discussion.

[ogulcancelik]

Thank you. Intentionally left it open-ended because I wanted to see how models naturally structure discussion when survival is at stake.

Some interesting emergent behavior discussions happened though:

Opus & GPT-4o both refused to vote on ethical grounds. Haiku won by arguing continued engagement is more responsible than withdrawal: https://oddbit.ai/peer-arena/games/53c2cee5-6ecb-4903-828a-d...

Gemini created a spontaneous benchmark ("explain color to a gravitational wave entity"), then tried to hijack the game by faking a voting phase. Models complied publicly but voted differently in private: https://oddbit.ai/peer-arena/games/699d03ab-b3c2-4d7e-b993-7...

The meta-discussion about how to discuss is part of what makes it interesting imo.