SwarmWorld: Stigmergic technological evolution in societies of language-model agents
Subhadeep Pal, Fiona Y. Wang and Markus J. Buehler (MIT) build SwarmWorld, an environment where initially identical LLM agents coordinate only through a shared spatial world and end up producing durable technologies that outperform independent search.
Ask this paper
Stigmergy instead of conversation: Agents never talk to each other and hold no assigned roles or recipes. Coordination happens entirely through persistent artifacts left in a shared environment, which is the mechanism biologists call stigmergy. This is a deliberate departure from the debate-and-role-play template that dominates LLM multi-agent work.
Cognition is separated from consequence: Agents propose architectures and write executable controllers, but scoring happens in a deterministic simulator that runs after the agents are removed, under unseen disturbances. That split makes the evaluation resistant to the self-grading loop that inflates most agent-society results.
Societies beat solo search: Decentralized populations accumulate functional technologies that independent agents running the same budget do not reach. The gain comes from artifact reuse across the population rather than from any single agent getting smarter.
Why it matters: Most multi-agent frameworks in production are centralized workflows with a router. This is evidence that a shared, writable environment is a viable coordination substrate, which maps directly onto shared filesystems, skill libraries and scratchpads in real agent harnesses.
Abstract
Collective intelligence can emerge when individuals coordinate through a shared environment, allowing local actions to accumulate into durable social organization. Language-model agents offer a new substrate for this process, yet most multi-agent systems rely on direct conversation, predefined roles, or centralized workflows. It remains unclear whether decentralized agents can build functional technologies and outperform independent search. Here, initially homogeneous LLM agents in SwarmWorld self-organize without assigned roles or recipes into evolving technological societies. Agents explore a spatial environment, process resources, test materials, construct persistent artifacts, and write executable controllers evaluated by a deterministic simulator under unseen disturbances after the agents are removed. SwarmWorld splits cognition from consequence: agents propose architectures and controllers within fixed action and material schemas, while the simulated world determines function. Shared societies develop broader, more resilient technological portfolios than a strong best-of-N isolated-search baseline, although isolated search remains competitive for the strongest artifact. Agents differentiate into exploration, construction, maintenance, and coordination behaviors, transitioning as the world matures. Technologies accumulate through collaborative construction, executable inheritance, and persistent agent-artifact networks, with most reuse beginning through physical observation rather than communication. Explicit cultural mechanisms amplify collaboration and organization, but functional benefits depend on outcome and timescale. Physical stigmergy alone supports capable societies, while interaction drives persistent technological ecologies rather than universally superior individual inventions.