Abstract page for arXiv paper 2611.10089: META-IR: Meta-Introspective Reasoning for Adaptive Safety Alignment in LLM