ChartX
收藏资源简介:
ChartX是由上海人工智能实验室构建的多模态评估集,涵盖18种图表类型、7种图表任务、22个学科主题,以及高质量的图表数据。该数据集旨在全面严格地评估现有MLLMs在图表领域的性能。ChartX数据集包含48,000个多模态图表数据,每个图表数据包括图像、逗号分隔值(CSV)、Python代码和文本描述四种模态。数据集的应用领域包括图表理解和生成,旨在解决图表数据的多模态理解和推理问题。
ChartX is a multimodal evaluation dataset constructed by the Shanghai AI Laboratory, covering 18 chart types, 7 chart-related tasks, 22 disciplinary topics, and high-quality chart data. This dataset is designed to comprehensively and rigorously evaluate the performance of existing multimodal large language models (MLLMs) in the chart domain. The ChartX dataset contains 48,000 multimodal chart samples, with each sample including four modalities: image, comma-separated values (CSV), Python code, and text description. Its application scenarios include chart understanding and generation, aiming to address the multimodal understanding and reasoning issues related to chart data.

- 1ChartX & ChartVLM: A Versatile Benchmark and Foundation Model for Complicated Chart Reasoning上海人工智能实验室 · 2024年



