arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2104.05232cs.CLcs.CYcs.LG
收藏

双重扰动:论稳健性与反事实偏差评价的稳健性

Double Perturbation: On the Robustness of Robustness and Counterfactual Bias Evaluation

UCLA(加州大学洛杉矶分校)

Chong Zhang, Jieyu Zhao, Huan Zhang, Kai-Wei Chang, Cho-Jui Hsieh

摘要:

稳健性和反事实偏差通常在测试数据集上进行评估。然而,这些评价是否可靠?如果测试数据集受到轻微扰动,评估结果是否保持不变?在本文中,我们提出了一个“双扰动”框架,以发现模型的弱点超出了测试数据集。该框架首先对测试数据集进行扰动,构造出大量与测试数据相似的自然句子,然后对单个词替换的预测变化进行诊断。我们应用这个框架来研究两种基于扰动的方法,用于分析模型的鲁棒性和反事实偏见的英语。(1)为了鲁棒性,我们专注于同义词替换,并确定易受攻击的例子,预测可以改变。我们提出的攻击在原始和经过鲁棒训练的CNN和Transformers上找到易受攻击的例子时,成功率很高(96.0%-99.8%)。(2)对于反事实偏见,我们专注于替换人口统计标记(例如,性别,种族),并测量预期预测在构建的句子中的偏移。我们的方法能够揭示隐藏的模型偏差,而不是直接显示在测试数据集中。我们的代码可在https://github.com/chong-z/nlp-second-order-attack上获得。

英文摘要:

Robustness and counterfactual bias are usually evaluated on a test dataset. However, are these evaluations robust? If the test dataset is perturbed slightly, will the evaluation results keep the same? In this paper, we propose a "double perturbation" framework to uncover model weaknesses beyond the test dataset. The framework first perturbs the test dataset to construct abundant natural sentences similar to the test data, and then diagnoses the prediction change regarding a single-word substitution. We apply this framework to study two perturbation-based approaches that are used to analyze models' robustness and counterfactual bias in English. (1) For robustness, we focus on synonym substitutions and identify vulnerable examples where prediction can be altered. Our proposed attack attains high success rates (96.0%-99.8%) in finding vulnerable examples on both original and robustly trained CNNs and Transformers. (2) For counterfactual bias, we focus on substituting demographic tokens (e.g., gender, race) and measure the shift of the expected prediction among constructed sentences. Our method is able to reveal the hidden model biases not directly shown in the test dataset. Our code is available at https://github.com/chong-z/nlp-second-order-attack.

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。

↑