Prioritizing interpretability in failure mode extraction is crucial for understanding model failures.