Reproduce “MusicSwarm: Biologically Inspired Intelligence for Music Composition” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Markus J. Buehler

MusicSwarm presents a decentralized swarm intelligence system for music composition that orchestrates multiple frozen GPT-5-mini agents through stigmergic coordination. This approach generates complex, long-form musical compositions, outperforming traditional multi-agent systems and single-shot methods in creative intelligence metrics, particularly for open-ended tasks.

Key insights

  • Creativity and specialized intelligence can emerge as a system-level property from the dynamic coordination of multiple frozen general-purpose foundation models, not solely from individual model parameter updates.
  • Biologically inspired principles, such as stigmergic coordination and iterative self-organization, effectively guide autonomous agents to produce complex, coherent creative outputs.
  • Decentralized multi-agent systems offer a compute- and data-efficient pathway to long-horizon creative tasks and emergent specialization by leveraging policy-over-prompts for adaptation.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.