To enhance decision-making in various industries, researchers at MIT and the MIT-IBM Computing Research Lab have developed ChartNet, a dataset designed to improve the performance of vision-language models (VLMs) in interpreting charts. This innovative dataset contains over one million diverse chart images, each accompanied by visual, numerical, and textual information to facilitate robust reasoning. The researchers employed a novel two-step synthetic data generation method, allowing smaller open-source models to significantly outperform larger commercial models in tasks such as data extraction and chart summarization. Jovana Kondic, a lead author of the study, emphasized that ChartNet serves as a comprehensive resource for AI practitioners, enabling effective chart understanding. The dataset also includes expert-annotated chart data, which can be used to fine-tune existing models for specific applications. The research highlights a critical gap in high-quality training data for VLMs, which has previously hindered accurate chart interpretation. The findings will be presented at the IEEE Computer Vision and Pattern Recognition Conference, with plans for future enhancements to ChartNet.