Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs [pdf](martins1612.github.io)159 points by tmnvdb 1 year ago | 88 comments
Screenshot pending…LLMs Demonstrate Behavioral Self-Awareness [pdf](martins1612.github.io)11 points by disparate4927 1 year ago | 1 comments
![Screenshot of martins1612.github.io: Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs [pdf]](https://hn.urlbox.com/shots/renders/2025/2/25/6b40cd1f-2485-4c17-b186-6f455059310a.jpg)