LLM Security Vulnerabilities

Papers on detecting, evaluating, and mitigating security threats in large language models, including prompt injection attacks, skill extraction, and safety failures through red-teaming and fuzzing approaches.