Gemini Pro vs. GPT-4V in Education: The study compares the classification performance of Gemini Pro and GPT-4V in educational settings, utilizing visual question answering techniques. Here are the key insights:

• Both models were tested on their ability to read text-based rubrics and automatically score student-drawn models in science education.
• The analysis included quantitative and qualitative methods, using NERIF (Notation-Enhanced Rubrics for Image Feedback) prompting.
• GPT-4V outperformed Gemini Pro significantly in terms of scoring accuracy and Quadratic Weighted Kappa—a measure of consistency.
• The study highlighted GPT-4V’s superior image processing and fine-grained text analysis capabilities.

Implications and Future Research:

• GPT-4V’s higher performance in handling complex multimodal tasks makes it a valuable tool for educational applications. The ability to integrate and interpret various types of data efficiently could revolutionize educational assessments.
• Further research could explore the potential of such AI models in other multimodal educational settings, potentially transforming the way educational content is delivered and assessed.