OrgAgent: Organize Your Multi-Agent System like a Company

Summary

The strongest pro-hierarchy counterweight to the self-organization and diversity-collapse results: a three-layer company-style framework (governance, execution, compliance) that beats flat collaboration across reasoning tasks while cutting tokens, reporting a 102.73% performance gain with 74.52% fewer tokens for GPT-OSS-120B on SQuAD 2.0. It also states the conditions under which hierarchy pays - stable skill assignment, controlled information flow, layered verification - which is a usable set of moderators. Cite as evidence that organizational structure shapes effectiveness, cost and coordination behaviour simultaneously, and distinguish on outcome: it measures task accuracy and token cost, not exploration, justification soundness or traceability.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-09-07 literature sweep as a medium-priority candidate for the Architecture as Epistemology (ISR) paper (angle: MAS topology / structure vs. aggregation rule; organization theory applied to AI agents; MAS failure and cost analyses). Verified against: Opened https://arxiv.org/abs/2604.01020 (arXiv abstract page; confirmed title, 6 authors, cs.MA, v1 1 Apr 2026, full abstract read)

Critique / open questions

Queued — not yet assessed.

Key quotes

While large language model-based multi-agent systems have shown strong potential for complex reasoning, how to effectively organize multiple agents remains an open question. (abstract)

In this paper, we introduce OrgAgent, a company-style hierarchical multi-agent framework that separates collaboration into governance, execution, and compliance layers. (abstract)