Pay Attention to Real World Perturbations! Natural Robustness Evaluation in Machine Reading Comprehension
This addresses the robustness of MRC models for real-world applications, highlighting a gap in current evaluation methods, but it is incremental as it builds on existing benchmarks and perturbation concepts.
The authors tackled the problem of evaluating machine reading comprehension (MRC) models' robustness to real-world textual perturbations by introducing a framework using Wikipedia edit history, and found that natural perturbations cause performance degradation in state-of-the-art models like Flan-T5 and LLMs, with training on perturbed examples improving robustness but leaving a gap compared to unperturbed data.
As neural language models achieve human-comparable performance on Machine Reading Comprehension (MRC) and see widespread adoption, ensuring their robustness in real-world scenarios has become increasingly important. Current robustness evaluation research, though, primarily develops synthetic perturbation methods, leaving unclear how well they reflect real life scenarios. Considering this, we present a framework to automatically examine MRC models on naturally occurring textual perturbations, by replacing paragraph in MRC benchmarks with their counterparts based on available Wikipedia edit history. Such perturbation type is natural as its design does not stem from an arteficial generative process, inherently distinct from the previously investigated synthetic approaches. In a large-scale study encompassing SQUAD datasets and various model architectures we observe that natural perturbations result in performance degradation in pre-trained encoder language models. More worryingly, these state-of-the-art Flan-T5 and Large Language Models (LLMs) inherit these errors. Further experiments demonstrate that our findings generalise to natural perturbations found in other more challenging MRC benchmarks. In an effort to mitigate these errors, we show that it is possible to improve the robustness to natural perturbations by training on naturally or synthetically perturbed examples, though a noticeable gap still remains compared to performance on unperturbed data.