Towards Understanding Jailbreak Attacks in LLMs: A Representation Space Analysis

Open in new window