Automatic Grading

Automatic grading uses machine learning, particularly large language models (LLMs) and deep learning architectures like convolutional neural networks and transformers, to automate the assessment of student work, aiming to reduce grading workload and improve consistency. Current research focuses on enhancing grading accuracy, particularly for open-ended questions and essays, by improving model robustness, incorporating human-in-the-loop methods, and developing explainable models to increase transparency and trust. This field holds significant potential for improving educational efficiency and fairness in large-scale assessments, while also presenting challenges related to model bias, security (e.g., vulnerability to adversarial attacks), and the need for careful evaluation methodologies.