Stop Uploading Test Data in Plain Text: Practical Strategies for Mitigating Data Contamination by Evaluation Benchmarks
Alon Jacovi11{}^{1}start_FLOATSUPERSCRIPT 1 end_FLOATSUPERSCRIPT Avi Caciularu1,212{}^{1,2}start_FLOATSUPERSCRIPT 1 , 2 end_FLOATSUPERSCRIPT Omer Goldman11{}^{1}start_FLOATSUPERSCRIPT 1 end_FLOATSUPERSCRIPT Yoav Goldberg1,313{}^{1,3}start_FLOATSUPERSCRIPT 1 , 3 end_FLOATSUPERSCRIPT 11{}^{1}start_FLOATSUPERSCRIPT 1 end_FLOATSUPERSCRIPT Bar Ilan University 22{}^{2}start_FLOATSUPERSCRIPT 2 end_FLOATSUPERSCRIPT Google Research 33{}^{3}start_FLOATSUPERSCRIPT 3 end_FLOATSUPERSCRIPT Allen Institute for Artificial Intelligence
Abstract
Data contamination has become prevalent and challenging with the rise of models pretrained on large automatically-crawled corpora. For closed models, the training data becomes a trade secret, and even for open models, it is not trivial to detect contamination. Strategies such as leaderboards with hidden answers, or using test data which is guaranteed to be unseen, are expensive and become fragile with time. Assuming that all relevant actors value clean test data and will cooperate to mitigate data contamination, what can be done? We propose three strategies that can make a difference: (1) Test data made public should be encrypted with a public key and licensed to disallow derivative distribution; (2) demand training exclusion controls from closed API holders, and protect your test data by refusing to evaluate without them; (3) avoid data which appears with its solution on the internet, and release the web-page context of internet-derived data along with the data. These strategies are practical and can be effective in preventing data contamination.
中文速览
大规模网络爬取语料让"数据污染"问题愈演愈烈——评测集在不知情的情况下混入训练数据，导致模型得分虚高、评测结论失真。作者系统梳理了这一问题的两条主要来源：一是自动爬取互联网形成的训练集，二是调用封闭API时泄露的测试数据，并指出现有的排行榜隐藏答案或持续创建新数据等补救手段既昂贵又治标不治本。为此，论文提出三条面向"善意参与者"的预防性策略：用公钥加密公开发布的测试集并附加禁止衍生分发的授权协议、拒绝为没有训练排除机制的封闭API提供评测数据、避免使用答案已随题目在网上公开的数据并在发布时附上原始网页上下文。这三条策略成本低、可操作性强，旨在从源头阻断污染而非事后检测，对维护NLP评测的可信度具有现实意义。
原文 arXiv:2305.10160;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2305.10160v2