The paper "Learning to Transform Dynamically for Better Adversarial Transferability" addresses the issue of adversarial examples, which are crafted by adding imperceptible perturbations to benign inputs to deceive neural networks. The authors introduce a novel approach called Learning to Transform (L2T) to enhance the adversarial transferability of generated samples across different models. L2T dynamically selects the optimal combination of transformations in each iteration to improve the diversity of transformed images, thereby enhancing the adversarial transferability. The selection of optimal transformation combinations is conceptualized as a trajectory optimization problem, and a reinforcement learning strategy is employed to solve it effectively. Extensive experiments on the ImageNet dataset and practical tests with Google Vision and GPT-4V demonstrate that L2T outperforms existing methods in enhancing adversarial transferability, confirming its effectiveness and practical significance. The code for L2T is available at <https://github.com/RongyiZhu/L2T>.The paper "Learning to Transform Dynamically for Better Adversarial Transferability" addresses the issue of adversarial examples, which are crafted by adding imperceptible perturbations to benign inputs to deceive neural networks. The authors introduce a novel approach called Learning to Transform (L2T) to enhance the adversarial transferability of generated samples across different models. L2T dynamically selects the optimal combination of transformations in each iteration to improve the diversity of transformed images, thereby enhancing the adversarial transferability. The selection of optimal transformation combinations is conceptualized as a trajectory optimization problem, and a reinforcement learning strategy is employed to solve it effectively. Extensive experiments on the ImageNet dataset and practical tests with Google Vision and GPT-4V demonstrate that L2T outperforms existing methods in enhancing adversarial transferability, confirming its effectiveness and practical significance. The code for L2T is available at <https://github.com/RongyiZhu/L2T>.