让大模型在虚拟电脑环境中训练，能激发通用智能并提升效率。
Computer Environments Elicit General Agentic Intelligence in LLMs
- 用最小化代码沙盒环境，让模型学会调用外部资源、文件管理和代码执行。
- 无需额外训练，强模型在多个领域性能提升最高达15.5%，耗 token 减少8倍。
- 弱模型通过仅在沙盒中训练即可掌握环境交互，适合构建通用智能体。
大语言模型的智能不仅依赖自身能力，还需与外部环境互动。当前将计算机赋予模型已成为主流趋势。然而，计算机环境本身的价值尚未系统研究，尤其其对通用能力的激发潜力。本文提出 LLM-in-Sandbox，将计算机虚拟为仅含基础功能的代码沙盒，证明此极简设置可激发模型在任务求解中的计算机元能力：外部资源访问、文件管理与代码执行。无需额外训练，强模型在数学、物理、化学、生物医学、长上下文理解及指令遵循等多个任务上性能提升最高达15.5%，同时令牌消耗减少最多8倍。此外，我们开发了 LLM-in-Sandbox-RL，仅使用非代理数据在沙盒中训练模型，使弱模型也能利用环境并内化交互能力。结果表明，计算机环境能激发通用智能，带来效率提升，并可通过训练被有效利用，是构建通用智能体的有力基础。
原文摘要 · Abstract (English)
Agentic intelligence in large language models (LLMs) requires not only model intrinsic capabilities but also interactions with external environments. Equipping LLMs with computers now represents a prevailing trend. However, the computer environment's intrinsic value has not been systematically investigated, particularly its potential to elicit general capabilities. Here we introduce LLM-in-Sandbox, which virtualizes the computer as a code sandbox with only basic functionalities, and demonstrate that this minimal setting elicits computer-based meta-capabilities for general task solving: external resource access, file management, and code execution. Without additional training, strong models achieve substantial gains (up to 15.5%) across mathematics, physics, chemistry, biomedicine, long-context understanding, and instruction following, while reducing token consumption by up to 8 times. Furthermore, we develop LLM-in-Sandbox-RL to train models exclusively on non-agentic data within the sandbox, empowering weaker models to harness the environment and internalize these interactions. Our results demonstrate that computer environments elicit general intelligence, yield efficiency gains, and can be harnessed through training, serving as a promising foundation for generalist agents.
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。