From Solo to Social Learning: Characterizing Recursive Social Improvement in LLMs

2026-09-29

AI Summary

This study investigates whether large language model (LLM) agents, while pursuing individual rewards, can achieve recursive social improvement through mutual learning. To this end, it introduces the concept of recursive social improvement and constructs a multi-agent simulation environment constrained by token budgets. By integrating skill-file revision with dynamic peer selection strategies, the work systematically examines how independent search, imitation, and interactive behaviors influence population-level performance. The findings reveal that although LLMs enhance efficiency by replicating peers’ skills, such replication does not yet improve ultimate effectiveness. Furthermore, observing peers optimizes resource allocation and accelerates skill discovery. Crucially, the results demonstrate that while current LLMs exhibit high social learning efficiency, their absolute performance remains inferior to that of independent learners.

Abstract

Large language models (LLMs) can now improve themselves by revising the instructions they follow, and LLM agents are increasingly orchestrated to work together on complex problems. However, self-improvement methods typically optimize one system at a time, and multi-agent frameworks often have every model work toward a shared goal. We ask a different question. When each agent pursues its own reward, can self-improving LLMs learn from one another well enough to improve the whole population? We call this capability recursive social improvement. We study populations that revise skill files and choose whether, when, and whom to copy from. Independent search, learning from peers, and acting all share one token budget. In controlled environments, established social-learning algorithms benefit from peers, but three LLMs do not. They earn less reward per token than solo learners, and explore too narrowly or run out of tokens before acting. We then let the models write and revise their own skills. Observing peers changes how they improve, helping one model find useful skills sooner and another spend less on private search. Neither, however, outperforms independent learners at the same cost. Skills are copied, revised, and passed on, so one discovery can seed further search. Yet these exchanges concentrate the population around fewer independent discoveries. Together, these results show that LLMs can make learning more efficient by copying from peers, but not yet more effective.

Problem

• Large Language Models
• Recursive Social Improvement
• Self-improvement
• Multi-agent Learning
• Social Learning

Innovation

• Recursive Social Improvement
• Large Language Models
• Multi-Agent Systems
• Social Learning
• Skill Revision