How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
[2023/10/14] We have updated the results on GPT-4V. The attack success rate is 45%!.
The main features of thu-ml/attack-bard are: Multimodal Attacks.
Projects with overlapping indexed features include: adversarial-for-goodness/co-attack — This is the official PyTorch implement of the paper "Towards Adversarial Attack on Vision-Language Pre-training… cgcl-codes/advclip — The implementation of our ACM MM 2023 paper "AdvCLIP: Downstream-agnostic Adversarial Examples in Multimodal… ericyinyzy/vlattack — VLAttack: Multimodal Adversarial Attacks on Vision-Language Tasks via Pre-trained Models. euanong/image-hijacks — This is the code for Image Hijacks: Adversarial Images can Control Generative Models at Runtime. haochen-luo/cropa — This repository contains the code and data for the paper "An Image Is Worth 1000 Lies: Transferability of Adversarial… kaiyuancui/ultrabreak — ICLR 2026.
The implementation of our ACM MM 2023 paper "AdvCLIP: Downstream-agnostic Adversarial Examples in Multimodal Contrastive Learning"
VLAttack: Multimodal Adversarial Attacks on Vision-Language Tasks via Pre-trained Models
This is the code for Image Hijacks: Adversarial Images can Control Generative Models at Runtime.
This is the official PyTorch implement of the paper "Towards Adversarial Attack on Vision-Language Pre-training Models" at ACM Multimedia 2022.