Improving LLM Outputs Against Jailbreak Attacks With Expert Model Integration

Improving LLM Outputs Against Jailbreak Attacks With Expert Model Integration

Using LLMs in a production environment presents security challenges that include vulnerabilities to jailbreaks and prompt injections, which can result in harmful outputs for humans or the enterprise. The challenge is amplified when working within a specific domain, as topics generally accepted for L...

Full description

Saved in:
Bibliographic Details
Main Authors: Tatia Tsmindashvili, Ana Kolkhidashvili, Dachi Kurtskhalia, Nino Maghlakelidze, Elene Mekvabishvili, Guram Dentoshvili, Orkhan Shamilov, Zaal Gachechiladze, Steven Saporta, David Dachi Choladze
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Expert model
generative AI
jailbreak attacks
large language models
prompt injections
Online Access:https://ieeexplore.ieee.org/document/11095693/
Tags: Add Tag
No Tags, Be the first to tag this record!

Internet

https://ieeexplore.ieee.org/document/11095693/

Similar Items