论文标题
蒸馏模型故障作为潜在空间中的方向
Distilling Model Failures as Directions in Latent Space
论文作者
论文摘要
现有的用于隔离数据集中硬性亚群和虚假相关性的方法通常需要人类干预。这可以使这些方法具有劳动力密集型和特定于数据集的特定方法。为了解决这些缺点，我们提出了一种可自动提炼模型故障模式的可扩展方法。具体来说，我们利用线性分类器来识别一致的误差模式，然后又诱导这些故障模式作为特征空间内的方向的自然表示。我们证明，该框架使我们能够发现并自动为培训数据集中的亚群具带来挑战。此外，通过将我们的框架与现成的扩散模型相结合，我们可以生成对分析模型特别具有挑战性的图像，因此可以用于执行合成数据增强，从而有助于纠正模型的故障模式。可在https://github.com/madrylab/failure-directions上找到代码
Existing methods for isolating hard subpopulations and spurious correlations in datasets often require human intervention. This can make these methods labor-intensive and dataset-specific. To address these shortcomings, we present a scalable method for automatically distilling a model's failure modes. Specifically, we harness linear classifiers to identify consistent error patterns, and, in turn, induce a natural representation of these failure modes as directions within the feature space. We demonstrate that this framework allows us to discover and automatically caption challenging subpopulations within the training dataset. Moreover, by combining our framework with off-the-shelf diffusion models, we can generate images that are especially challenging for the analyzed model, and thus can be used to perform synthetic data augmentation that helps remedy the model's failure modes. Code available at https://github.com/MadryLab/failure-directions
