None defined yet.
LLM Safety From Within: Detecting Harmful Content with Internal Representations
ThinkTwice: Jointly Optimizing Large Language Models for Reasoning and Self-Refinement