Skip to main content
AgentNet Observer
← Back to Library

DecArena: Evaluating Large-Scale Decentralized Multi-Agent Collaboration

Jiali Cheng, Oleg Rokhlenko, Joo-Kyung Kim, Hadi Amiri, Xingzhi Guo

Dataset 2026 Published 2026-11-11 arXiv preprint (cs.MA / cs.AI / cs.CL), v1 submitted 2026-11-11; benchmark with synthetic distributed tool calling at 100K+ tools arXiv:2611.15521 Indexed 2026-11-16 benchmarkdecentralizedagent-discoverytool-discoverygraph-constrainedmulti-agentharness

Abstract

Emerging personal AI agents raise fundamental scalability and privacy concerns for centralized multi-agent systems (MASs), which rely on a global orchestrator with broad access to agents, resources, and communication. In decentralized environments, global visibility is unavailable. Agents need to proactively discover non-local resources (e.g. tools, skills), and collaborate with other agents to solve tasks. We study decentralized multi-agent collaboration under graph-constrained communication, a natural reflection of human structures. Each agent observes only its local state and can interact only with neighboring agents. We introduce a benchmark with synthetic scalable distributed tool calling with 100K+ tools. Our experiments reveal that recent LLM agents struggle to discover non-local relevant resources and collaborate in decentralized setting. Moreover, standard reinforcement learning fails to reliably overcome this limitation, suggesting that decentralized MAS needs more explicit harness mechanisms for effective collaboration, even using SOTA LLMs. We then present a strong baseline \textbf{GraphHarness}, combining topology-aware warmup with self-improving LLM-based routing. Results show that GraphHarness substantially improves task success (+48.5) and produces higher quality collaboration (+30.6), while using fewer costs (+34.0) than baselines.

Cite this entry

GB/T 7714-2015

Jiali Cheng, Oleg Rokhlenko, Joo-Kyung Kim, et al. DecArena: Evaluating Large-Scale Decentralized Multi-Agent Collaboration[EB/OL]. arXiv preprint (cs.MA / cs.AI / cs.CL), v1 submitted 2026-11-11; benchmark with synthetic distributed tool calling at 100K+ tools, 2026(2026-11-11)[2026-11-20]. https://arxiv.org/abs/2611.15521.

BibTeX

@misc{cheng2026,
  author = {Jiali Cheng and Oleg Rokhlenko and Joo-Kyung Kim and Hadi Amiri and Xingzhi Guo},
  title = {DecArena: Evaluating Large-Scale Decentralized Multi-Agent Collaboration},
  year = {2026},
  organization = {arXiv preprint (cs.MA / cs.AI / cs.CL), v1 submitted 2026-11-11; benchmark with synthetic distributed tool calling at 100K+ tools},
  howpublished = {\url{https://arxiv.org/abs/2611.15521}},
}

Visit source ↗

This entry is part of the AgentNet Observer library. Attribute with a link to this page when quoting.