Fetching the paper…

Latent Adversarial Training Improves Robustness to Persistent Harmful Behaviors in LLMs · Around