---
title: "University of Washington Finds LLM Agent Swarms Copy Well but Explore Poorly"
subtitle: "A new study tests whether populations of self-improving LLM agents actually benefit from watching each other, and finds copying crowds out exploration."
url: "https://alphasignal.ai/news/university-of-washington-finds-llm-agent-swarms-copy-well-but-explore-poorly"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-09-29T20:37:10.000Z"
updated: "2026-10-03T12:24:25.093Z"
type: "Paper"
topics:
  - "Agents"
  - "Post Training"
subtopics:
  - "Multi Agent"
  - "Fine Tuning"
  - "Rl"
upvotes: 146
source_url: "https://arxiv.org/abs/2609.38516"
comments: 0
truncated: true
---

# University of Washington Finds LLM Agent Swarms Copy Well but Explore Poorly

> A new study tests whether populations of self-improving LLM agents actually benefit from watching each other, and finds copying crowds out exploration.

# LLM agent populations copy well and explore poorly

A University of Washington preprint finds that individually rewarded LLM agents can copy and revise peers’ skills. Those exchanges failed to beat independent learners when both groups received the same total token budget. In controlled bandit tasks, social agents earned less reward per token; in open-ended skill evolution, peer observation sometimes accelerated progress while final performance remained level. For agent-swarm developers, the results show how communication costs and early convergence can erase the benefits of sharing.

Self-improving agents usually optimize their own prompts, tools, or reusable instructions. Multi-agent systems often coordinate agents around one shared objective. The study examines a different arrangement: each agent has its own task and reward, can inspect peers’ work, and must allocate limited compute among private search, observation, and execution.

## A social-learning test with a real budget

The paper, [available as a preprint](https://arxiv.org/abs/2609.38516), calls this process **recursive social improvement**. An agent develops a procedure, another agent copies it, and later agents revise the copied version. Useful discoveries can therefore propagate and become starting points for further search.

Each round forces an agent to choose among searching privately, observing a peer, and acting on its task. All three choices consume the same finite token budget. A token spent reading another agent’s logs or skill file cannot be spent generating an answer.

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [From Solo to Social Learning: Characterizing Recursive Social Improvement in LLMs](https://arxiv.org/abs/2609.38516)
- [From Solo to Social Learning (HTML)](https://arxiv.org/html/2609.38516v1)
- [OpenEvolve](https://github.com/codelion/openevolve)