What is it about?
Large Language Models (LLMs) show promise in automating assessment scoring and feedback via prompting. It is common practice to add assessment rubrics to the prompt, but so far this has been done in a naive and brute-force way: by simply copying and pasting the rubric into the prompt. We show how to create better prompts by using the rubric in more intelligent ways, which significantly increase the performance of LLMs.
Featured Image
Photo by Isaac Smith on Unsplash
Why is it important?
Grading assessments is a very time-consuming task. AI approaches such as Large Language Models show promise in helping alleviating costs, but they need to show acceptable performance in order to be useful in educational settings. Our work helps closing this gap.
Perspectives
This is a work in progress. While our initial experiments are very promising, it is still unclear if they generalise to other kinds of assessments and how useful the feedback is. This is something we plan to investigate in future work. Nevertheless, the method highlights the need to look at problems in detail when proposing AI-related solutions. We can only go so far with brute force approaches: smart and robust adoption of AI solutions will require tailoring solutions in the best way, with the help of domain experts.
Daniel Beck
RMIT University
Read the Original
This page is a summary of: Rubric-guided Prompting for Automatic Assessment Scoring, June 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3774398.3811581.
You can read the full text:
Resources
Contributors
The following have contributed to this page
