BenchMarker：一种受教育启发的工具包，用于突出多项选择基准测试中的缺陷
BenchMarker: An Education-Inspired Toolkit for Highlighting Flaws in Multiple-Choice Benchmarks
AI总结：
本文提出BenchMarker工具，利用LLM检测多项选择基准测试中的三项常见缺陷，通过人工标注和审计12个基准测试，发现自动和众包数据中存在大量缺陷，且修复方法可能引入新问题，影响NLP评估质量。
摘要：
多项选择问答（MCQA）在NLP中是标准形式，但基准测试缺乏严格质量控制。我们提出了BenchMarker，一种受教育启发的工具包，利用LLM法官标记三项常见MCQ缺陷：1）污染：问题恰好在线出现；2）捷径：选项中的提示使猜测成为可能；3）书写错误：基于19条教育评分标准的结构性或语法问题。我们通过人工标注验证BenchMarker，然后运行工具审核12个基准测试，发现：1）MCQA基准测试中存在缺陷，尤其是自动和众包数据——我们检测到47%的TruthfulQA问题出现在网络上，100%的HellaSwag违反了多个书写规则；2）污染的MCQ倾向于提高准确性，而书写错误会降低准确性并改变排名；3）先前的基准修复方法针对特定问题（如用LLM生成的干扰项降低准确性），但无意中引入了新问题（如不合理的干扰项，多个正确答案）。总体而言，MCQ中的缺陷会降低NLP评估质量，但教育研究提供了解决方案。我们发布BenchMarker以弥合这两个领域并改进MCQA基准测试设计。
英文摘要：
Multiple-choice question answering (MCQA) is standard in NLP, but benchmarks lack rigorous quality control. We present BenchMarker, an education-inspired toolkit using LLM judges to flag three common MCQ flaws: 1) contamination: items appearing exactly online; 2) shortcuts: cues in the choices that enable guessing; and 3) writing errors: structural/grammatical issues based on a 19-rule education rubric. We validate BenchMarker with human annotations, then run the tool to audit 12 benchmarks, revealing: 1) flaws persist in MCQA benchmarks, especially automatically-made and crowdsourced data - we detect 47% of TruthfulQA appears online and 100% of HellaSwag violates multiple writing rules; 2) contaminated MCQs tend to inflate accuracy, while writing errors tend to lower it and change rankings beyond random; and 3) prior benchmark repairs address their targeted issues (i.e., lowering accuracy with LLM-written distractors), but inadvertently add new flaws (i.e. implausible distractors, many correct answers). Overall, flaws in MCQs degrade NLP evaluation, but education research offers a path forward. We release BenchMarker to bridge the fields and improve MCQA benchmark design.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
