BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//OneJoy//events//EN
BEGIN:VEVENT
UID:onejoy-event-35773@onejoy
DTSTAMP:20261027T224428Z
DTSTART:20260916T013000Z
DTEND:20260916T033000Z
SUMMARY:Andrey presents "Refusal geometry reflects refusal training"
LOCATION:
DESCRIPTION:Refusal geometry reflects refusal training: diverse refusal prefixes can raise stable rank and weaken refusal vector ablation attacks\n\nAbstract\n\nRefusal training protects AI models from jailbreaks by training models to decline unsafe queries\, reducing the risk of misuse. Recent work finds that refusal behavior in aligned language models can be mediated by a single activation direction or a low-dimensional refusal subspace shared across harmful prompts: ablating those directions suppresses refusa\nhttps://onejoy.io/event/35773
URL:https://onejoy.io/event/35773
END:VEVENT
END:VCALENDAR