原文整理页

探讨 AI 评估(Evals)的辨别力属性,强调有效的评估应能区分出具有实质差异的系统

来源作者:Madhu Guru (@realmadhuguru)原始来源:https://x.com/realmadhuguru/status/2092058332735693264

中文导读

探讨 AI 评估(Evals)的辨别力属性,强调有效的评估应能区分出具有实质差异的系统。

正文 Markdown

How to build great evals - part 8. The discriminatory property of evals. A hill-climbing eval is useful when it can separate AI systems that are meaningfully different. Imagine you run an eval on five AI systems: A: 94 B: 93 C: 95 D: 94 E: 92 Now suppose you already know A