Generating Valid and Natural Adversarial Examples with Large Language Models | IEEE Conference Publication | IEEE Xplore