Effective Robustness against Natural Distribution Shifts for Models with Different Training Data
This work addresses a methodological issue in robustness evaluation for machine learning models, particularly when comparing across diverse training datasets, and is incremental in refining existing metrics.
The paper tackles the problem of evaluating effective robustness for models trained on different data distributions by proposing a new metric that uses multiple in-distribution test sets, revealing that prior gains for zero-shot CLIP-like models diminish under this improved evaluation.
"Effective robustness" measures the extra out-of-distribution (OOD) robustness beyond what can be predicted from the in-distribution (ID) performance. Existing effective robustness evaluations typically use a single test set such as ImageNet to evaluate the ID accuracy. This becomes problematic when evaluating models trained on different data distributions, e.g., comparing models trained on ImageNet vs. zero-shot language-image pre-trained models trained on LAION. In this paper, we propose a new evaluation metric to evaluate and compare the effective robustness of models trained on different data. To do this, we control for the accuracy on multiple ID test sets that cover the training distributions for all the evaluated models. Our new evaluation metric provides a better estimate of effective robustness when there are models with different training data. It may also explain the surprising effective robustness gains of zero-shot CLIP-like models exhibited in prior works that used ImageNet as the only ID test set, while the gains diminish under our new evaluation. Additional artifacts including interactive visualizations are provided at https://shizhouxing.github.io/effective-robustness.