知识卡片
判分器也需要反例样本
判分样例应覆盖等价表达、真正不等、无效语法、缺失、嵌套和边界语义;同时检查误判为对与误判为错。原书 16 个演示通过只证明这些案例的行为,不能证明通用数学判等;新增反例还应说明期待结果的任务依据。
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
- · 评估推理模型
原书出处 · 1 处
- · 查看原书依据
Exercise 3.1: Adding more test cases Try to think of additional test cases, ideally challenging ones, and add them to the run_demos_table() function. Can you find cases where the check fails incorrectly? With the grade_function implemented, we now have the building blocks in place to evaluate the LLM. In the next section, we will load a math dataset on which we will evaluate the LLM.
参考来源
- 位置:《Build a Reasoning Model (From Scratch)》第03章《评估推理模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Build a Reasoning Model (From Scratch)/_epub-src/OEBPS/Text/chapter-3.html)
- 结论依据:摘录中关于判分器也需要反例样本的定义、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:Exercise 3.1: Adding more test cases Try to think of additional test cases, ideally challenging ones, and add them to the run_demos_table() function. Can you find cases where the check fails incorrectly? With the grade_function implemented, we now have the building blocks in place to evaluate the LLM. In the next section, we will load a math dataset on which we will evaluate the LLM.