Double Visual Defense
- Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)
This is the official code for the paper "Double Visual Defense: Adversarial Pre-training and Instruction Tuning for Improving Vision-Language Model Robustness". This code can be used to produce vision language models (VLMs), like LLaVA, with enhanced robustness to adversarial attacks (e.g. jailbreaks).
- Short Name / Acronym:
- DVD
- Site Accession Number:
- LLNL-CODE-2002980
- Software Type:
- Scientific
- License(s):
- MIT License
- Research Organization:
- Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- USDOE National Nuclear Security Administration (NNSA)Primary Award/Contract Number:AC52-07NA27344
- DOE Contract Number:
- AC52-07NA27344
- Code ID:
- 154189
- OSTI ID:
- code-154189
- Country of Origin:
- United States
Similar Records
Swap Path Network for Robust Person Search Pre-training
CLIP Representation Enhanced Predicate Estimation
Adversarial Robustness Limits
Software
·
Thu Mar 20 20:00:00 EDT 2025
·
OSTI ID:code-154933
CLIP Representation Enhanced Predicate Estimation
Software
·
Thu Jul 20 20:00:00 EDT 2023
·
OSTI ID:code-128865
Adversarial Robustness Limits
Software
·
Mon May 20 20:00:00 EDT 2024
·
OSTI ID:code-134198