We present an overview of the Colliding With Adversaries challenge on robust learning in High Energy Physics, held at ECML PKDD 2025. The challenge was split into two tasks: (1) generating adversarial examples to attack a deep learning model, and (2) developing a model robust to unseen adversarial attacks while maintaining strong performance on clean data. Both tasks used a tabular dataset derived from two simulated CMS LHC collision processes, featuring 87 continuous features per sample. We summarize the challenge setup, evaluation metrics, dataset characteristics, and submission statistics. Finally, we describe the methods of the winning team, Team APTH, who employed a white-box gradient-based attack for the Adversarial Attack Task and a method-aware robustness strategy using an ensemble of feature-embedding models for the Robust Model Task. The challenge highlights the potential of adversarial deep learning in high-energy physics, promotes research on robust models for tabular data, and seeks to introduce non-physicists to the rich datasets typical of high-energy physics experiments.