作者位次 #4/7ACL 2026Poster· Natural Language Processing
Why LLM Safety Guardrails Collapse After Fine-tuning: A Similarity Analysis Between Alignment and Fine-tuning Datasets
Lei Hsiung · Tianyu Pang · Yung-Chen Tang · Linyue Song · Tsung-Yi Ho · Pin-Yu Chen · Yaoqing Yang
摘要
Recent advancements in large language models (LLMs) have underscored their vulnerability to safety alignment jailbreaks, particularly when subjected to downstream fine-tuning. However, existing mitigation strategies primarily focus on reactively addressing jailbreak incidents after safety guardra...