Aligning LLMs at inference time by suppressing internal concepts

2 points | by adebayoj 8 hours ago