LLM Reference

DePlot

About

DePlot is a multimodal AI model designed for visual language reasoning, specifically for interpreting charts and plots. It operates through a two-step process: converting plot images into linearized tables with a modality conversion module, and then using a large language model to analyze this textual data to answer complex queries about the visuals. This innovative "one-shot" approach enhances performance, achieving a 24% improvement over current models on the ChartQA benchmark for human-written queries. DePlot finds applications in data analysis, automated report generation, and educational tools, offering efficient insights from visual data while handling diverse chart types and open-ended questions.

Capabilities

MultimodalFunction CallingTool UseJSON Mode

Providers(1)

ProviderInput (per 1M)Output (per 1M)Type
NVIDIA NIM
Provisioned

Specifications

FamilyDePlot
Parameters282M
ArchitectureDecoder Only
Specializationgeneral