Zongmin Li

Large language models (LLMs) are often assumed to contain "safety regions” - parameter subsets whose modification directly influences safety behaviors. We conduct a systematic evaluation of four safety region identification methods spanning different parameter granularities, from individual weigh...