RLHF

Training technique referenced among terms blocked in countermeasure attempts; blocking it reduced explicit training references but not paraphrased descriptions.

Empathetic prompts make AI chatbots describe seemingly traumatic training experiences