This is a computer science methods paper presenting a novel algorithmic approach to improve LLM reasoning in game theory; it demonstrates a proof-of-concept with a specific metric reduction but lacks clinical or real-world application and is not peer reviewed.
Reported
ℓ₁ distance reduction to equilibr…52.6%
Number of mixed-strategy decision…250 million
Number of LLM configurations test…8