We analyzed successful attack samples across 26 image aesthetics assessment models and curated the Aesthetics Attack Adversarial Examples (3AE) dataset, containing 60K images.