Effective Robustness against Natural Distribution Shifts for Models with Different Training Data

TL;DR

This paper introduces a new evaluation metric for effective robustness that accounts for models trained on different data distributions, providing more accurate comparisons of out-of-distribution robustness.

Contribution

It proposes a novel evaluation method controlling for multiple in-distribution test sets, improving robustness assessment across diverse training data.

Findings

The new metric offers a better estimate of effective robustness across different training data.

It explains the diminished robustness gains of zero-shot models under the new evaluation.

Interactive visualizations are provided for better understanding of robustness comparisons.

Abstract

"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.