Leveraging BART to Assess CS1 C++ Programming Assignments using Rubric-based Criteria
Learn how to use BART for automated grading of C++ programming assignments with rubric-based criteria, improving grade predictions to match instructor behavior
- Preprocess C++ programming assignments into unified sequences for transformer models
- Fine-tune BART using rubric-aware, multitask learning to improve grade predictions
- Evaluate the performance of the fine-tuned BART model against general-purpose LLMs and instructor grading behavior
- Apply the trained model to automate grading of new student submissions
- Compare the grade predictions from the BART model with those from instructors to refine the assessment process
This research benefits educators and instructors in computer science, particularly those teaching introductory programming courses, as it provides a method for automated grading that aligns with instructor assessment criteria. Developers of educational tools can also apply this knowledge to create more accurate grading systems.
💡 Rubric-aware fine-tuning of transformer models like BART can produce more accurate grade predictions that reflect instructor grading behavior
Automate grading of C++ assignments with BART & rubric-based criteria 🤖💻
Key Takeaways
Learn how to use BART for automated grading of C++ programming assignments with rubric-based criteria, improving grade predictions to match instructor behavior
Full Article
Abstract:
arXiv:2606.03814v1 Announce Type: new Abstract: This paper investigates rubric-aware, multitask fine-tuning of transformer models for automated grading of introductory C++ programming assignments, with the goal of producing grade predictions that better reflect instructor grading behavior than general-purpose LLMs. Using multi-semester CS1 data, student submissions are paired with numeric scores, letter-grade buckets, and assignment rubrics, then preprocessed into unified sequences for transform
DeepCamp AI