ChartGPT-Llama3: Natural Language to Chart Generation

ChartGPT-Llama3 is an 8 billion parameter language model, fine-tuned from Meta-Llama-3-8B-Instruct, specifically engineered for generating charts directly from natural language instructions. Developed by yuan-tian, this model addresses the challenge of translating abstract user requests into concrete data visualizations.

Key Capabilities

• Natural Language Chart Generation: Interprets natural language prompts to produce detailed chart specifications.
• Step-by-Step Output: Generates responses in a structured, multi-step format, including column selection, data filtering, aggregation, chart type, encodings, and sorting.
• Data-Agnostic Input: Designed to work with various tabular data inputs, requiring table name, headers, header types, and data examples.
• Research-Backed: Based on the research presented in the paper "ChartGPT: Leveraging LLMs to Generate Charts from Abstract Natural Language" published in IEEE Transactions on Visualization and Computer Graphics.

Training Details

The model was fine-tuned on the chartgpt-dataset-llama3 dataset, building upon the robust capabilities of the Meta-Llama-3-8B-Instruct base model.

Ideal Use Cases

This model is particularly well-suited for applications requiring automated data visualization from user queries, enabling non-technical users to create charts without needing to write code or understand complex visualization libraries. It can be integrated into data analysis platforms, business intelligence tools, or interactive dashboards to enhance user experience and accessibility for data exploration.