Utilizing Vision Language Models (VLMs) for Efficient and Objective Student Assessment

Abstract

This study explores the application of Vision Language Models (VLMs) in evaluating student work, focusing on their potential to enhance efficiency and objectivity in assessment processes. VLMs, integrating natural language processing and computer vision, offer a novel approach to analyzing student responses, particularly in assignments involving visual elements. This paper outlines the functionality of VLMs, discusses their advantages and limitations, and provides practical guidance on their implementation. It also includes examples of prompt engineering and showcases initial results from a pilot study conducted at SMP N 32 Padang, demonstrating the potential of VLMs in a real-world educational setting. This method allows teacher to asses written text by the students on assignments that are also presented in written or image format. The use of VLM is expected to further develop efficiency and precision in student assessment.