ChartNet: A Million-Scale, High-Quality Multimodal Dataset for Robust Chart Understanding

2026-03-27

AI Summary

Existing vision-language models struggle to jointly and effectively comprehend the geometric structures, numerical data, and natural language associated with charts. To address this limitation, this work proposes a code-guided synthetic approach to construct the largest open-source multimodal chart understanding dataset to date, comprising 1.5 million high-quality samples spanning 24 chart types and six plotting libraries. The dataset features fine-grained quintuple alignment among chart images, generation code, underlying data tables, natural language summaries, and question-answer pairs. Rigorous quality control ensures reliability, including expert-annotated, real-world, safety-focused, and localization-specific subsets. Fine-tuned vision-language models leveraging this dataset demonstrate substantial performance gains across multiple benchmarks, validating its efficacy as a strong supervisory signal and advancing the development of robust, generalizable foundation models for visual understanding.

Abstract

Understanding charts requires models to jointly reason over geometric visual patterns, structured numerical data, and natural language -- a capability where current vision-language models (VLMs) remain limited. We introduce ChartNet, a high-quality, million-scale multimodal dataset designed to advance chart interpretation and reasoning. ChartNet leverages a novel code-guided synthesis pipeline to generate 1.5 million diverse chart samples spanning 24 chart types and 6 plotting libraries. Each sample consists of five aligned components: plotting code, rendered chart image, data table, natural language summary, and question-answering with reasoning, providing fine-grained cross-modal alignment. To capture the full spectrum of chart comprehension, ChartNet additionally includes specialized subsets encompassing human annotated data, real-world data, safety, and grounding. Moreover, a rigorous quality-filtering pipeline ensures visual fidelity, semantic accuracy, and diversity across chart representations. Fine-tuning on ChartNet consistently improves results across benchmarks, demonstrating its utility as large-scale supervision for multimodal models. As the largest open-source dataset of its kind, ChartNet aims to support the development of foundation models with robust and generalizable capabilities for data visualization understanding. The dataset is publicly available at https://huggingface.co/datasets/ibm-granite/ChartNet

Problem

• chart understanding
• vision-language models
• multimodal dataset
• data visualization
• cross-modal alignment

Innovation

• code-guided synthesis
• multimodal alignment
• chart understanding
• quality filtering
• foundation models