Scientific Chart
Scientific chart understanding is a rapidly evolving field focused on enabling computers to interpret and reason about the information contained within charts, diagrams, and figures. Current research emphasizes the development and refinement of multimodal large language models (MLLMs), often leveraging techniques like visual instruction tuning and pre-training with raw data values to improve accuracy in extracting and interpreting numerical and relational information. This work aims to enhance scientific communication and data analysis by automating tasks such as chart captioning, question answering, and text role classification within charts, ultimately improving accessibility and efficiency in scientific research.
Papers
September 30, 2024
July 19, 2024
June 24, 2024
February 8, 2024