misk@piefed.social to Technology@lemmy.zipEnglish · 2 months agoOne long sentence is all it takes to make LLMs misbehavewww.theregister.comexternal-linkmessage-square7linkfedilinkarrow-up153arrow-down11 cross-posted to: technology@lemmy.ml
arrow-up152arrow-down1external-linkOne long sentence is all it takes to make LLMs misbehavewww.theregister.commisk@piefed.social to Technology@lemmy.zipEnglish · 2 months agomessage-square7linkfedilink cross-posted to: technology@lemmy.ml
minus-squareEvotech@lemmy.worldlinkfedilinkEnglisharrow-up1·2 months agoThis refers spesifically to local models like llama 70b Not that cloud models don’t have this issue, ut they very much have defence in depth for this type of attack
This refers spesifically to local models like llama 70b
Not that cloud models don’t have this issue, ut they very much have defence in depth for this type of attack