Jailbreaking Frontier Foundation Models Through Intention Deception
Large (vision-)language models exhibit remarkable capability but remain highly susceptible to jailbreaking. Existing saf
#cs.CL
共收录 424 条相关情报。
← 返回所有主题An Information-Geometric Framework for Stability Analysis of Large Language Models under Entropic Stress
As large language models (LLMs) are increasingly deployed in high-stakes and operational settings, evaluation strategies
SSG: Logit-Balanced Vocabulary Partitioning for LLM Watermarking
提出SSG方法，通过对数几率平衡的词汇分区提升LLM水印在低熵场景（代码生成、数学推理）下的检测能力。
Behavioral Canaries: Auditing Private Retrieved Context Usage in RL Fine-Tuning
提出行为金丝雀审计机制，通过在偏好数据中注入触发器-奖励信号，检测RL微调是否使用了受保护检索数据。