Summary of Light-weight Fine-tuning Method For Defending Adversarial Noise in Pre-trained Medical Vision-language Models, by Xu Han et al.
Light-weight Fine-tuning Method for Defending Adversarial Noise in Pre-trained Medical Vision-Language Models
by Xu Han, Linghao Jin, Xuezhe Ma, Xiaofeng Liu
First submitted to arxiv on: 2 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: 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 This paper investigates how pre-trained Vision-Language Models (VLMs) handle noisy data in medical applications. By crafting noisy datasets using multi-modal adversarial attacks, researchers found that moderate noise improves model robustness and transferability but excessive noise hinders downstream task performance. To address this issue, the authors propose a rectify adversarial noise (RAN) framework to defend against adversarial attacks and mitigate upstream noise during fine-tuning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper explores how pre-trained models handle noisy data in medical applications. Researchers created noisy datasets using special attacks to test if their findings would be affected by mistakes. Surprisingly, they found that a little bit of noise made the model better at doing its job, but too much noise was bad for performance. To fix this problem, scientists developed a new way called RAN to protect against these attacks and remove the noise before using the model. |
Keywords
» Artificial intelligence » Fine tuning » Multi modal » Transferability
