OfficeTown 框架：用于专业组织任务模拟与 Qwen3.5-9B 模型评测
办公 agent 别测孤立工单：MSRA OfficeTown，406 任务 / 4836 准则；Astra Pass@1 74.4%，Qwen GDPval 65.75→73.65
研究人员介绍了 OfficeTown 框架，该框架用于模拟专业组织中的任务和上下文，并通过在 OfficeTown 数据上进行监督微调和在线蒸馏，显著提升了 Qwen3.5-9B 模型在专业工作基准上的表现。
该研究提出了 OfficeTown 框架，用于模拟专业组织中的任务和上下文，并展示了其在提升 Qwen3.5-9B 模型在专业工作基准上的表现方面的价值。
Authors:Hongchen Wei, Daohong Ying, Zhi Hong, Aidan Wu, Bei Liu, Ruichun Ma, Yifan Yang, Qi Dai, Kai Qiu, Rui Xu, Dongdong Chen, Zhenzhong Chen, Chong Luo
Abstract:AI agents increasingly use files and software tools to produce professional deliverables. Such work draws on materials, decisions, and feedback accumulated across assignments. We introduce OfficeTown, a generative simulation framework for professional organizations in which tasks and their context arise from continuing organizational work. From a natural-language description of an organization's purpose and activities, OfficeTown initializes an organization whose members coordinate work, delegate bounded assignments to assistant agents, and review their deliverables. Artifacts, decisions, and unresolved matters persist and shape subsequent work, while responsibilities and permissions determine what each member can see and do. We curate work requests, their context, and review feedback into executable resources for evaluation and learning. We construct OfficeTown Bench with 406 tasks and 4,836 evaluation criteria from 23 simulated organizations across eight professional domains, incorporating domain-expert feedback. Frontier-model evaluations reveal substantial remaining headroom. Supervised fine-tuning and on-policy distillation on OfficeTown data substantially improve Qwen3.5-9B on external professional-work benchmarks, including GDPval. Supplementary evaluations show that the principal--assistant harness achieves higher scores than the single- and multi-agent baselines tested. These results highlight the value of organizational simulation in generating work-grounded data and environments for the continued development of generalist agents.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.23194 [cs.AI] |
| (or arXiv:2610.23194v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.23194 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Hongchen Wei [view email]
[v1]
Fri, 16 Oct 2026 09:32:16 UTC (6,114 KB)
来源：Agent Control Flash News · arxiv.org
Cursor Agent 架构全景拆解：从影子工作区到多文件推演引擎
全面剖析 Cursor 在 Agent 模式下的底层技术实现，解析影子分支隔离、LSP 上下文剪枝以及预测性补全的核心架构细节与工程实践经验。
Vercel v0 推出全新多 Agent 协作画布与实时全栈部署预览
动态速评：v0 正在从单一组件生成器演进为全栈应用搭建协作台，尤其是与 Neon/Postgres 的直连极大降低了全栈原型验证成本。