TY  - RPRT
TI  - An Interpretable N-gram Perplexity Threat Model for Large Language Model Jailbreaks
AU  - Valentyn Boreiko
AU  - Alexander Panfilov
AU  - Vaclav Voracek
AU  - Matthias Hein
AU  - Jonas Geiping
PY  - 2025
UR  - https://arxiv.org/abs/2410.16222
ID  - 2410.16222
ER  -