评分 6 · 来源：cs.AI updates on arXiv.org · 发布于 2026-04-14
评分依据：Calibrates VLM safety perception via lightweight injection without fine-tuning cost or utility degradation. Leverages inherent VLM recognition of unsafe content.
Risk Awareness Injection: Calibrating Vision-Language Models for Safety without Compromising Utility
发布
采集
学术前沿 6.0 分
— Calibrates VLM safety perception via lightweight injection without fine-tuning cost or utility degradation. Leverages inherent VLM recognition of unsafe content.