Summary of Feedback-based Modal Mutual Search For Attacking Vision-language Pre-training Models, by Renhua Ding et al.
Feedback-based Modal Mutual Search for Attacking Vision-Language Pre-training Models
by Renhua Ding, Xinze Zhang, Xiao Yang, Kun He
First submitted to arxiv on: 27 Aug 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed Feedback-based Modal Mutual Search (FMMS) attack paradigm leverages target model feedback to iteratively refine adversarial examples, driving them into the adversarial region. By introducing a novel modal mutual loss (MML), FMMS aims to push away matched image-text pairs while drawing mismatched pairs closer in feature space, guiding the update directions of adversarial examples. This approach significantly outperforms state-of-the-art baselines on Flickr30K and MSCOCO datasets for image-text matching tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary FMMS is a new way to make VLP models weak by creating fake images or text that look like they’re from another model, but are actually trying to trick the model. This happens because some bad guys found ways to make attacks work better on these models. To stop this, FMMS uses a special loss function and feedback from the target model to create more clever fake examples that can fool VLP models. This makes it harder for attackers to use these models in real life. |
Keywords
* Artificial intelligence * Loss function
