Agensh: Scaling Organizational Intelligence to 1,024 Agents
- ID: 6191e86b
- 原文链接: https://arxiv.org/abs/2609.26781
- PDF: https://arxiv.org/pdf/2609.26781v1
- 作者: Zhihao Zhan, Ting Song, Li Dong, Shaohan Huang, Jianxun Lian, Yan Xia, Furu Wei
- 发布日期: 2026-09-22
- 更新日期: 2026-09-22
- 分类: cs.CL
- 来源类型: paper
- 标签: arxiv, multi-agent, orchestration, scalability, paper
- 质量评分: 4/5
- 抓取时间: 2026-09-24T04:23:22Z
中文导读
多 agent 系统可并行降低复杂任务延迟，但现有 harness 受限于中央 orchestrator 的任务分配与协调容量Agensh 去掉中央 orchestrator，让并发 worker 自组织跑合作循环：持续收集上下文认领并自派子任务执行共享发现异步验证并合并进度；底层由共享工作区消息接口和共享上下文三个组件支撑在 ProgramBench 最难 5 个任务上用 GPT-5.6-sol(high) 评测：agent 从 1 扩到 128，平均最终测试通过率从 19.31% 升至 28.78%（相对提升约 49%）；pandoc 任务上从 1 扩到 1024 个 agent，通过率从 33.89% 升至 55.06%轨迹分析显示自组织合作形态随组织规模扩大逐渐涌现并标准化
为什么值得关注
去中心化的千人 agent 组织：自组织合作循环替代中央 orchestrator，pandoc 任务 1024 agent 通过率 33.9%55.1%
关键信息
- 论文标题：Agensh: Scaling Organizational Intelligence to 1,024 Agents
- 作者：Zhihao Zhan, Ting Song, Li Dong, Shaohan Huang, Jianxun Lian, Yan Xia, Furu Wei
- arXiv：https://arxiv.org/abs/2609.26781
- 发布时间：2026-09-22
- 更新时间：2026-09-22
- arXiv 分类：cs.CL, cs.MA
- 可选注释：13 pages, 6 figures
- 关联标签：arxiv, multi-agent, orchestration, scalability, paper
English Abstract
A multi-agent system can reduce latency on complex tasks by executing work concurrently. Several pioneering harness frameworks support multi-agent systems. However, the scalability of current multi-agent harnesses is often constrained by a central orchestrator's capacity to allocate tasks and coordinate workers. To address this limitation, we introduce Agensh, a scalable self-organized multi-agent harness without a central orchestrator: concurrent workers execute a multi-agent cooperation loop, continuously gathering context, claiming and self-assigning sub-tasks, taking action and sharing findings, verifying results, and merging progress in an asynchronous manner. The loop is supported by the agentic organization infrastructure comprising three components: a shared workspace holds proposed, ongoing, and completed work; a message interface lets workers communicate; and shared context retains reusable findings and work intentions. To test the scalability of Agensh, we evaluate it on the five hardest ProgramBench tasks with GPT-5.6-sol (high). Scaling from 1 to 128 agents raises the mean final test-pass rate from 19.31% to 28.78%, an approximately 49% relative improvement. Larger organizations reach comparable test-pass rates earlier. On pandoc, scaling from 1 to 1,024 agents raises the final test-pass rate from 33.89% to 55.06%. Worker trajectories further show that different forms of self-organized cooperation gradually emerges and standardizes as the organization grows. These results reveal the number of agents as a new scaling dimension for multi-agent organizations to expand the frontier of general intelligence, offering a practical solution for complex tasks under hard latency constraints or time budgets.
English Summary
Multi-agent systems can reduce latency on complex tasks, but current harnesses are constrained by a central orchestrator's capacity to allocate tasks and coordinate workers. Agensh is a scalable self-organized multi-agent harness without a central orchestrator: concurrent workers run a cooperation loop that gathers context, claims and self-assigns sub-tasks, acts, shares findings, verifies results, and merges progress asynchronously, supported by a shared workspace, a message interface, and shared context. On the five hardest ProgramBench tasks with GPT-5.6-sol (high), scaling from 1 to 128 agents raises mean final test-pass rate from 19.31% to 28.78%; on pandoc, scaling from 1 to 1,024 agents raises it from 33.89% to 55.06%. Worker trajectories show self-organized cooperation gradually emerging and standardizing as the organization grows.
Obsidian Notes
- 内容由
opencli arxiv paper拉取 arXiv 元数据与摘要生成。 - 中文导读与价值判断均锚定在条目已有摘要、论文摘要、作者、日期与分类信息上；未补充论文摘要之外的实验细节。
- 本页面属于 content-fetcher 能动下的高分筻补东生成，不修改原条目字段，仅产出可读的中英双语页面。