#machine learning
Nov 2025
Learning Robust Social Strategies with Large Language Models
A recent opponent-learning awareness algorithm is adapted to fine-tune LLMs toward multi-agent cooperation and non-exploitability, and a group-relative baseline is introduced that simplifies advantage computation in iterated games, enabling multi-agent training at LLM scale.
Dereck Piche, Mohammed Muqeeth, Milad Aghajohari et al.
· arXiv.org · 2 citations