用大模型实现自然语言操控表格，三代理系统提升准确率与可靠性。
SheetMind: An End-to-End LLM-Powered Multi-Agent Framework for Spreadsheet Automation
- 三代理协作：分解指令、生成结构化命令、验证意图一致性。
- 在221项任务中执行成功率100%，功能正确率达54.8%，优于基线。
- 可嵌入Google Sheets，适合需要自动化办公的用户使用。
我们提出SheetMind，一个基于大语言模型（LLM）的模块化多智能体框架，通过自然语言指令实现电子表格自动化。该框架采用分层代理系统，包含三个专用智能体：经理代理负责将复杂指令拆解为子任务；动作代理利用巴科斯-诺尔范式（BNF）语法将子任务转化为结构化命令；反思代理则验证生成动作与用户原始意图的一致性。我们在221项任务的SheetCopilot基准上使用GPT-3.5-Turbo评估了SheetMind，实现了100%的执行成功率和54.8%的功能正确率，优于基线方法（44.3%），同时保持了完美的执行可靠性。此外，我们在独立构建的数据集上进行消融实验，确认完整三代理配置始终优于所有简化变体。最后，我们将系统集成至Google Sheets，通过Workspace扩展提供服务。
原文摘要 · Abstract (English)
We present SheetMind, a modular multi-agent framework powered by large language models (LLMs) for spreadsheet automation via natural language instructions. In this paper, we introduce a hierarchical agentic system consisting of three specialized agents: Manager Agent that decomposes complex user instructions into subtasks; an Action Agent that translates these into structured commands using a Backus-Naur Form (BNF) grammar; and a Reflection Agent that validates alignment between generated actions and the user's original intent. We evaluate SheetMind on the 221-task SheetCopilot Benchmark with GPT-3.5-Turbo. SheetMind achieved 100% execution success and 54.8% functional correctness, exceeding SheetCopilot (44.3%) while maintaining perfect execution reliability. We also conduct ablation study on a separately curated dataset to confirm that the full three-agent configuration consistently outperforms all partial variants. Lastly, we integrate our system into Google Sheets via a Workspace extension.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。