METR：Research（网页）·· 12 小时前精选AI 评分78
METR 提出"任务时长"指标：AI 可完成任务长度每约 7 个月翻倍
Measuring AI Ability to Complete Long Software Tasks March 19, 2025 We propose measuring AI performance in terms of the *length* of tasks AI agents can complete. We show that this metric has been consistently exponentially increasing over the past 6 years, with a doubling time of around 7 months. Extrapolating this trend predicts that, in under a decade, we will see AI agents that can independently complete a large fraction of software tasks that currently take humans days or weeks. Read more
AI 导读
METR 提出用任务长度（按人类专家耗时衡量）来度量 AI 能力，发现 6 年来模型能以 50% 可靠性自主完成的任务长度呈指数增长，翻倍时间约 7 个月。当前模型对耗时 4 分钟以内任务成功率近 100%，超过约 4 小时则不足 10%；Claude 3.7 Sonnet 的 50% 成功率时间视界约一小时。
推荐理由
METR 用人类任务时长衡量智能体能力，给出 6 年约 7 个月翻倍的指数趋势及其外推含义。
来源：METR：Research（网页） · metr.org