fp.
newest
Open in hackernews
Aligning LLMs at inference time by suppressing internal concepts
https://www.guidelabs.ai/post/steerling-8b-alignment-without-retraining/
2
•
adebayoj
•
2h ago