Summary of Rethinking Comprehensive Benchmark For Chart Understanding: a Perspective From Scientific Literature, by Lingdong Shen et al.
Rethinking Comprehensive Benchmark for Chart Understanding: A Perspective from Scientific Literature
by Lingdong Shen, Qigqi, Kun Ding, Gaofeng Meng, Shiming Xiang
First submitted to arxiv on: 11 Dec 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Computer Vision and Pattern Recognition (cs.CV); 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 introduces a new benchmark for evaluating multimodal models’ ability to understand complex scientific charts. The existing benchmarks have limitations, such as a narrow range of chart types and overly simplistic visual elements, which can lead to inflated performance scores that don’t hold up in real-world scenarios. To address these challenges, the authors introduce Scientific Chart QA (SCI-CQA), which focuses on flowcharts as a critical category. They curated a dataset of 202,760 image-text pairs from top-tier computer science conferences and refined it to 37,607 high-quality charts with contextual information. The benchmark also includes a novel evaluation framework inspired by human exams, featuring both objective and open-ended questions. Furthermore, the authors propose an efficient annotation pipeline that reduces data annotation costs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about creating a new way to test how well computers understand complex scientific diagrams. Right now, we don’t have good ways to do this because most testing involves simple charts and easy questions. But real-world scientists use many different types of charts and ask harder questions. To solve this problem, the authors made a big dataset of 200,000+ image-text pairs from top computer science conferences. They then filtered it down to 37,000 high-quality diagrams with extra information. The new benchmark has a special way of testing that’s like how humans take exams. It also helps make annotating data easier and cheaper. |
