MIT researchers teach AI models to interpret charts | MIT News
MIT researchers teach AI models to interpret charts | MIT News
https://news.mit.edu/2026/mit-researchers-teach-ai-models-to-interpret-charts-0603
Publish Date: 2026-06-03 00:00:00
Source Domain: news.mit.edu
- Researchers from MIT and the MIT-IBM Computing Research Lab developed ChartNet, a dataset that includes over a million charted data points to train vision-language models (VLMs) to interpret charts effectively.
- ChartNet includes not only chart images but also the code used to generate them, textual descriptions, tables with numerical information, and questions with answers to teach models how to answer chart-related questions.
- The dataset enables open-source VLM models to surpass larger commercial models in data extraction and chart summarization tasks, potentially allowing small firms to utilize AI more affordably.
- Using synthetic data generation methods, ChartNet allows researchers to effectively interpret complex multimodal data within charts, a crucial task for businesses across industries.
- Researchers plan on expanding the dataset further to include more complex data and gather feedback from the research community.
