newsNature Machine IntelligenceTrust 88 · LabPublished 1mo agoLive · 1mo ago
Capable language models can outgrow the benefits of collaboration
Nature Machine Intelligence, Published online: 24 July 2026; doi:10.1038/s42256-026-01268-y A controlled study of large language model agents across 260 configurations shows when multi-agent collaboration helps or hurts performance, and introduces a predictive model that selects the best architecture in 87% of held-out within-domain configurations.
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- PossiblePossibly related (embedding) · 68%Agents in the Wild: Where Research Meets Deployment →
- PossiblePossibly related (embedding) · 62%MECoBench: A Systematic Study of Multimodal Agent Collaboration in Embodied Environments →
- PossiblePossibly related (embedding) · 62%Conversable Complexity: Agentic LLM Collectives as Interpretable Substrates →
- PossiblePossibly related (embedding) · 61%PolyWorkBench: Benchmarking Multilingual Long-Horizon LLM Agents →
- PossiblePossibly related (embedding) · 61%Gloriaameng/Awesome-Agent-Harness →
- PossiblePossibly related (embedding) · 59%Palmyra x6 Technical Report: An Agentic, Tool-Use Model Post-Trained via Anchored Supervised Fine-Tuning →
- PossiblePossibly related (embedding) · 51%The Interaction Tax: When Communication Erases Diversity in Multi-Agent Teams →
- PossiblePossibly related (embedding) · 51%Future-House/aviary →
Covers
paperAgents in the Wild: Where Research Meets DeploymentpaperMECoBench: A Systematic Study of Multimodal Agent Collaboration in Embodied EnvironmentspaperConversable Complexity: Agentic LLM Collectives as Interpretable SubstratespaperPolyWorkBench: Benchmarking Multilingual Long-Horizon LLM AgentsrepoGloriaameng/Awesome-Agent-Harness
Covers (incoming)
paperPalmyra x6 Technical Report: An Agentic, Tool-Use Model Post-Trained via Anchored Supervised Fine-TuningpaperThe Interaction Tax: When Communication Erases Diversity in Multi-Agent TeamsrepoFuture-House/aviarypaperSwarmBench: Can Large Language Models Act as Agent Swarm Orchestrators?paperMURANO: Design, Run, and Reproduce Mechanistic Interpretability Experiments as Composable PipelinespaperSkill Issue: Are Skills Language-Invariant in LLMs?paperSwarmWorld: Stigmergic technological evolution in societies of language-model agentspaperWhat Do CAE Simulation Agents Really Need Beyond a Generic Harness?
Related across the graph
paperPolyWorkBench: Benchmarking Multilingual Long-Horizon LLM AgentspaperThe Interaction Tax: When Communication Erases Diversity in Multi-Agent TeamsrepoGloriaameng/Awesome-Agent-HarnesspaperPalmyra x6 Technical Report: An Agentic, Tool-Use Model Post-Trained via Anchored Supervised Fine-TuningrepoFuture-House/aviarypaperMECoBench: A Systematic Study of Multimodal Agent Collaboration in Embodied EnvironmentspaperMURANO: Design, Run, and Reproduce Mechanistic Interpretability Experiments as Composable PipelinespaperSwarmBench: Can Large Language Models Act as Agent Swarm Orchestrators?paperWhat Do CAE Simulation Agents Really Need Beyond a Generic Harness?paperSwarmWorld: Stigmergic technological evolution in societies of language-model agentspaperConversable Complexity: Agentic LLM Collectives as Interpretable SubstratespaperAgents in the Wild: Where Research Meets DeploymentpaperSkill Issue: Are Skills Language-Invariant in LLMs?