知识卡片
符号解析失败需要单列
生成文本可能不符合数学解析器语法。捕获已知异常能使评估继续,但若统一折算为错误,最终准确率就包含解析失败的影响。除布尔得分外宜记录抽取失败、语法不支持、判等未决和答案错误,便于人工抽样定位。
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
The sympy_parser takes into account what looks like an excessive amount of error cases, but these are all errors that I encountered when evaluating the model on all 500 MATH-500 problems, as the model does not always generate perfectly formatted outputs.
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:摘录中关于符号解析失败需要单列的定义、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 sympy_parser takes into account what looks like an excessive amount of error cases, but these are all errors that I encountered when evaluating the model on all 500 MATH-500 problems, as the model does not always generate perfectly formatted outputs.