Source |
Dark Reading |
Identifiant |
8420004 |
Date de publication |
2023-12-07 20:52:00 (vue: 2023-12-07 21:08:08) |
Titre |
Les chercheurs montrent comment utiliser un LLM pour jancher un autre Researchers Show How to Use One LLM to Jailbreak Another |
Texte |
"Tree of Attacks with Troping" est le dernier d'une série croissante de méthodes pour susciter un comportement involontaire d'un modèle grand langage.
"Tree of Attacks With Pruning" is the latest in a growing string of methods for eliciting unintended behavior from a large language model. |
Envoyé |
Oui |
Condensat |
another attacks behavior eliciting from growing how jailbreak language large latest llm methods model one pruning researchers show string tree unintended use |
Tags |
|
Stories |
|
Notes |
★★★
|
Move |
|