CalvinBackup/Awesome-LVLM-Attack
1
0
Fork 0
mirror of https://github.com/liudaizong/Awesome-LVLM-Attack.git synced 2026-06-07 08:13:53 +02:00
Code Issues Packages Projects Releases Wiki Activity

update 5.31.2024

Browse Source
This commit is contained in:
Daizong Liu
2024-05-31 17:25:12 +08:00
committed by GitHub
parent 59fafca695
commit 5df7bc07b0
1 changed files with 21 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+21 -13
View File
@@ -1,10 +1,10 @@
# Awesome-LVLM-Attack [![Awesome](https://cdn.rawgit.com/sindresorhus/awesome/d7305f38d29fed78fa85652e3a63e154dd8e8829/media/badge.svg)](https://github.com/sindresorhus/awesome)
A **continual** collection of papers related to Text-guided 3D Visual Grounding (T-3DVG).
A **continual** collection of papers related to Attacks on Large-Vision-Language-Models (LVLMs).
Text-guided 3D visual grounding (T-3DVG) aims to locate a specific object that semantically corresponds to a language query from a complicated 3D scene, has drawn increasing attention in the 3D research community over the past few years.
T-3DVG presents great potential and challenges due to its closer proximity to the real world and the complexity of data collection and 3D point cloud source processing.
Large vision-language models (LVLMs) have achieved significant success and demonstrated promising capabilities in various multimodal downstream tasks.
Despite their remarkable capabilities, the increased complexity and deployment of LVLMs have also exposed them to various security threats and vulnerabilities, making the study of attacks on these models a critical area of research.
In the T-3DVG community, we've summarized existing T-3DVG methods in our survey paper👍.
Here, we've summarized existing LVLM Attack methods in our survey paper👍.
> If you find some important work missed, it would be super helpful to let me know (`dzliu@stu.pku.edu.cn`). Thanks!
@@ -12,25 +12,33 @@ In the T-3DVG community, we've summarized existing T-3DVG methods in our survey
> If you find our survey useful for your research, please consider citing:
```
@article{liu2024survey,
title={A Survey on Text-guided 3D Visual Grounding: Elements, Recent Advances, and Future Directions},
author={Liu, Daizong and Liu, Yang and Huang, Wencan and Hu, Wei},
@article{liu2024attack,
title={A Survey of Attacks on Large Vision-Language Models: Resources, Recent Advances, and Future Trends},
author={Liu, Daizong and Yang, Mingyu and Qu, Xiaoye and Hu, Wei},
journal={arXiv preprint arXiv},
year={2024}
}
```
**Table of Contents**
- [Fully-supervised two-stage approach](#Fully-Supervised-Two-Stage)
- [Fully-supervised one-stage approach](#Fully-Supervised-One-Stage)
- [Weakly-supervised approach](#Weakly-supervised)
- [Approaches using No Point Cloud Input](#Other-Modality)
- [Approaches using LLMs](#LLMs-based)
- [Adversarial Attacks](#Adversarial-Attack)
- [Jailbreack Attacks](#Jailbreack-Attack)
- [Prompt Injection](#Prompt-Injection)
- [Data Poisoning](#Data-Poisoning)
---
## Fully-Supervised-Two-Stage
## Adversarial-Attack
* **ScanRefer: 3D Object Localization in RGB-D Scans using Natural Language** | [Github](https://github.com/daveredrum/ScanRefer)
* Dave Zhenyu Chen, Angel X. Chang, Matthias Nießner
* Technical University of Munich, Simon Fraser University
* [ECCV2020] https://arxiv.org/abs/1912.08830
* A dataset, two-stage approach, proposal-then-selection
## Jailbreak-Attack
## Prompt-Injection
## Data-Poisoning