The article titled "Probabilistic Robustness for Free? Revisiting Training via a Benchmark," published on November 4, 2025, explores the concept of probabilistic robustness in deep learning models. Probabilistic robustness refers to the ability of a model to maintain accurate predictions when subjected to random perturbations, offering a different perspective from traditional adversarial robustness. While adversarial robustness focuses on defending against worst-case, often deliberately crafted perturbations, probabilistic robustness measures performance under stochastic variations. This distinction underscores the importance of evaluating model reliability not only against targeted attacks but also in more general, probabilistic settings. The article emphasizes the relevance of probabilistic robustness as a complementary metric for assessing model robustness. By revisiting training methods through a benchmark, the study aims to shed light on how probabilistic robustness can be achieved or improved. This approach may provide insights into enhancing model reliability in practical scenarios where random noise is prevalent.