一作ICML 2026Poster
Position: Generative Distributional Integrity against Backdoor Attacks
Shuaibiao Han · Ruiyang Ni · Zhiguo Yang · Changlong Li · Perley Xu · Wenjie Ruan
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Shuaibiao Han · Ruiyang Ni · Zhiguo Yang · Changlong Li · Perley Xu · Wenjie Ruan
Ruiyang Ni · Changlong Li · Shuaibiao Han · Zhiyu Yi · Perley Xu · Wenjie Ruan
State-of-the-art large language models (LLMs) have achieved impressive results on various tasks. However, these architectures are vulnerable to jailbreak attacks, such as GCG and AutoDAN. Several defense strategies have been proposed to protect LLMs from generating harmful content, with most meth...
Shuaibiao Han · Ruiyang Ni · Zhiyu Yi · Changlong Li · Perley Xu · Wenjie Ruan
Although Large Language Models undergo rigorous safety alignment, they remain vulnerable to adversarial attacks. Existing methods, particularly gradient-based prompt optimization, suffer from high computational costs and produce uninterpretable, high-perplexity inputs. While recent logit-space at...