arXiv：cs.CL· Burak Agachan, Max van Duijn, Amirhossein Zohrehvand·· 3 小时前AI 评分58
arXiv 论文：LLM 智能体团队中扁平结构报告质量优于层级结构
Loop-Back Authority in LLM Agent Teams: A Paired Experiment on Flat and Hierarchical Coordination
AI 导读
arXiv 论文（arXiv:2609.14767）通过 43 组配对笔记本产品、86 次运行的商业智能报告实验发现，扁平团队报告在 Utility（d=0.42, p=0.009）和 Writing Clarity（d=0.34, p=0.030）上得分更高。
正文
Abstract:Does authority in AI teams improve the outcome? Organizational theory asserts that authority facilitates decision making, improving quality. Meanwhile, some nascent AI research suggests that revision under authority makes LLM output worse. Multi-agent LLM frameworks default to giving a Manager agent the authority to send a worker's output back for revision. Prior comparisons test the effect of authority using verifiable tasks. We conduct an experiment on an open-ended task, business-intelligence reporting, using a sample of 43 paired laptop products and 86 runs. Each report is written once by a hierarchical team and once by a flat team. We find that flat teams produce higher-quality reports, scoring higher on Utility (d = 0.42, p = 0.009) and Writing Clarity (d = 0.34, p = 0.030). The reports are the same length, but hierarchical team reports use 53% more hedging words such as "may" and "could", and each revision is associated with a 0.14-point drop in Writing Clarity on a 1 to 5 scale. Before any revision, the hierarchical team's first draft is indistinguishable from the flat team's report. In other words, the quality gap can be traced to revision. Authority improves quality when the Manager can verify the work, else when it can only provide feedback it has a negative effect on quality.
| Comments: | 8 pages, 3 figures, 3 tables, plus 21 pages of supplementary material. Code: this https URL |
| Subjects: | Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); General Economics (econ.GN) |
| Cite as: | arXiv:2609.14767 [cs.MA] |
| (or arXiv:2609.14767v2 [cs.MA] for this version) | |
| https://doi.org/10.48550/arXiv.2609.14767 arXiv-issued DOI via DataCite |
Submission history
From: Amirhossein Zohrehvand [view email]
[v1]
Sun, 13 Sep 2026 19:58:54 UTC (696 KB)
[v2]
Wed, 7 Oct 2026 17:51:49 UTC (696 KB)
来源：arXiv：cs.CL · arxiv.org