Abstract
Multimodal large language models (MLLMs) are increasingly used to interpret visualizations, yet current evaluations remain largely chart-centric and provide limited evidence of understanding of scientific visualization (SciVis). This study benchmarks six MLLMs on the scientific visualization literacy assessment test, a standardized SciVis literacy assessment comprising 49 items based on 18 scientific visualizations and illustrations, spanning 8 techniques and 11 task types.
We evaluate three closed-source and three open-source models under a closed-world protocol and compare their performance using data from 485 human participants. Results show that current MLLMs do not exhibit uniform SciVis literacy. Gemini is the strongest model overall, exceeding the human mean across the evaluated subsets, while the open-source models remain below the human baseline. Performance is highly uneven across techniques and tasks: models perform best on scientific illustration, search, and spatial understanding, but struggle on texture-based and integration-based visualizations and on quantitative estimation.
Error analysis reveals recurring failures in fine-grained quantitative estimation, flow-direction interpretation, and grounded encoding interpretation. These findings position SciVis literacy as a necessary benchmark dimension for evaluating multimodal AI systems. Our code and model outputs are publicly available at GitHub.
Blogger's Review: This study highlights the limitations of multimodal large language models in the realm of scientific visualization, particularly in quantitative analysis and comprehension of complex visual information. It provides critical directions for future model improvements and emphasizes the necessity of SciVis literacy as a standard for evaluation. The shortcomings of open-source models also indicate a need for further research and development.