Summary of Spa-vl: a Comprehensive Safety Preference Alignment Dataset For Vision Language Model, by Yongting Zhang et al.
SPA-VL: A Comprehensive Safety Preference Alignment Dataset for Vision Language Model
by Yongting Zhang, Lu Chen, Guodong Zheng, Yifeng Gao, Rui Zheng, Jinlan Fu, Zhenfei Yin, Senjie Jin, Yu Qiao, Xuanjing Huang, Feng Zhao, Tao Gui, Jing Shao
First submitted to arxiv on: 17 Jun 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
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 paper proposes the Safety Preference Alignment (SPA-VL) dataset for Vision Language Models (VLMs), which aims to address the challenges of aligning these models with safety preferences. SPA-VL is a large-scale, high-quality dataset that covers 6 harmfulness domains, 13 categories, and 53 subcategories, containing 100,788 samples of quadruples (question, image, chosen response, rejected response). The responses are collected from 12 open-source and closed-source VLMs to ensure diversity. Models trained with alignment techniques on the SPA-VL dataset exhibit substantial improvements in harmlessness and helpfulness while maintaining core capabilities. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a special set of data for computer programs that understand both words and pictures, called Vision Language Models (VLMs). These models need to be safe and help people, but they’re not always. The paper wants to make sure these models are safer and more helpful by creating a big dataset with many examples of questions, images, answers, and rejected answers. This dataset will help train the models to be safer and better at helping people. |
Keywords
» Artificial intelligence » Alignment
