Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training | Read Paper on Bytez