知识卡片
比例结论要检查分母和筛选条件
本例211项中58线性、45平坦、27次线性,三类共130项;45/130约35%,而45/211约21%。筛选这些类别不表示它们是涌现能力,更不能把平坦直接解释成随机水平;应先核对列定义与原始实验。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- · 查看原书依据
The code snippet gives you an alphabetical sampling of the 130 nonemergent capabilities cataloged by Google researchers. The “flat” labels mean that increasing the size of an LLM did not increase the accuracy of the LLM on these tasks by any measurable or statistically significant amount. You can see that 35% (45/130) of the nonemergent capabilities were labeled as having flat scaling. Sublinear scaling means
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第10章《真实世界中的大语言模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-10.html)
- 结论依据:摘录中关于比例结论要检查分母和筛选条件的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:The code snippet gives you an alphabetical sampling of the 130 nonemergent capabilities cataloged by Google researchers. The “flat” labels mean that increasing the size of an LLM did not increase the accuracy of the LLM on these tasks by any measurable or statistically significant amount. You can see that 35% (45/130) of the nonemergent capabilities were labeled as having flat scaling. Sublinear scaling means