Efficient Adversarial Training in LLMs with Continuous Attacks | Read Paper on Bytez