arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2402.02823cs.LGcs.AIcs.CLcs.CR
收藏

规避语言模型的数据污染检测(太)容易了

Evading Data Contamination Detection for Language Models is (too) Easy

ETH Zurich(苏黎世联邦理工学院)

Jasper Dekoninck, Mark Niklas Müller, Maximilian Baader, Marc Fischer, Martin Vechev

AI总结:

本文指出恶意模型提供商可故意制造数据污染以规避检测,并提出分类法揭示现有检测方法的漏洞,同时设计了EAL技术,能显著虚增基准性能且完全规避当前检测。

摘要:

大型语言模型被广泛使用,它们在基准测试上的表现经常引导用户偏好选择某一模型。然而,这些模型训练所用的海量数据可能会无意中导致与公开基准测试发生数据污染,从而损害性能测量的准确性。尽管近期开发的污染检测方法试图解决这一问题,但它们忽略了恶意模型提供商为规避检测而故意制造污染的可能性。我们认为这一设定至关重要,因为它引发了对公开基准测试可靠性的怀疑。为了更严谨地研究这一问题,我们提出了一种针对模型提供商和污染检测方法的分类法。这揭示了现有方法中的漏洞,我们利用EAL(一种简单却有效的污染技术)来利用这些漏洞,该技术能显著虚增基准测试性能,同时完全规避当前的检测方法。

英文摘要:

Large language models are widespread, with their performance on benchmarks frequently guiding user preferences for one model over another. However, the vast amount of data these models are trained on can inadvertently lead to contamination with public benchmarks, thus compromising performance measurements. While recently developed contamination detection methods try to address this issue, they overlook the possibility of deliberate contamination by malicious model providers aiming to evade detection. We argue that this setting is of crucial importance as it casts doubt on the reliability of public benchmarks. To more rigorously study this issue, we propose a categorization of both model providers and contamination detection methods. This reveals vulnerabilities in existing methods that we exploit with EAL, a simple yet effective contamination technique that significantly inflates benchmark performance while completely evading current detection methods.

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。

↑