In one-shot self-play without communication, two frontier LLMs beat the worst symmetric Nash baseline in several two-player game archetypes, while open-weight models vary and team coordination collapses for most models when action spaces grow.
Subjectivity and correlation in randomized strategies,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.MA 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Do LLMs Beat Nash? Testing Decentralized Coordination in Self-Play Multi-Agent Games
In one-shot self-play without communication, two frontier LLMs beat the worst symmetric Nash baseline in several two-player game archetypes, while open-weight models vary and team coordination collapses for most models when action spaces grow.