Accelerating Accurate Assignment Authoring Using Solution-Generated Autograders

Research from arXiv proposes 'solution-generated autograding' as an improved method for creating autograders in programming education. The current process of authoring autograders, often reliant on manual test case enumeration, is described as time-consuming and prone to inaccuracies, hindering the creation of sufficient practice problems for programming students. The new approach aims to provide a faster, more accurate, and more enjoyable method for autograder creation.

Why it matters

The proposed innovation in autograder generation directly addresses efficiency and accuracy in educational technology development, particularly in programming instruction. Improving the speed and reliability of autograder creation can significantly enhance the quantity and quality of learning resources available to students, fostering more effective skill development at scale.

Key insights

• Students learning to program require access to numerous practice problems.
• Autograders are crucial for providing rapid feedback on programming submissions.
• Current autograder authoring methods, particularly test case enumeration, are tedious and can result in inaccurate autograders.
• Inaccurate autograders misclassify student submissions.
• The slow pace of authoring accurate autograders limits the availability of practice problems.
• A new approach, 'solution-generated autograding,' is presented to accelerate and improve autograder creation accuracy and user experience.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.06572