📄 中文摘要
Anthropic最近发布了一项关于代理自主性的研究，深入分析了其自有的METR数据集。该研究旨在探讨人工智能代理在自主决策过程中的表现和能力，特别是在复杂环境下的适应性。通过对METR数据的详细分析，研究揭示了代理在面对不同任务时的决策策略和效率。这一研究为理解AI代理的自主性提供了新的视角，并为未来的AI系统设计提供了重要的参考。
Anthropic的代理自主性研究
出处: [AINews] Anthropic's Agent Autonomy study
发布: 2026年2月19日
📄 English Summary
[AINews] Anthropic's Agent Autonomy study
Anthropic has recently released a study on agent autonomy, providing an in-depth analysis of its proprietary METR dataset. The study aims to explore the performance and capabilities of AI agents in autonomous decision-making processes, particularly their adaptability in complex environments. Through a detailed examination of the METR data, the research reveals the decision-making strategies and efficiencies of agents when faced with various tasks. This study offers new insights into the autonomy of AI agents and serves as an important reference for the design of future AI systems.