JAB: Java Academic Benchmark

This repository hosts the Java Academic Benchmark (JAB) introduced in our paper, "Java Academic Benchmark: Exam-Based Evaluation of LLMs on Object-Oriented Programming."

Overview

JAB is the first exam-based benchmark specifically designed to rigorously evaluate LLMs on Java Object-Oriented Programming (OOP) capabilities.

JAB comprises 103 real Java exams administered between 2014 and 2024 at a top-tier academic institution, designed and curated by Professor Mirko Viroli, an experienced full professor and co-author of this project. This ensures that each task reflects pedagogically grounded and conceptually meaningful challenges.

Unlike benchmarks that focus on isolated code snippets, JAB features full exam-style problems with expert-authored JUnit test suites, targeting class- and interface-level reasoning. Designed to assess human learners, these problems isolate common misconceptions and test deep understanding, offering granular insights into where and how models fall short.

Resources

• 📄 Paper: https://www.sciencedirect.com/science/article/pii/S0164121226002669
• 💻 Source Code & Evaluation: https://github.com/disi-unibo-nlp/jab

Dataset Structure

Each exam in the JAB dataset is structured as follows:

• year: Year of the exam session
• session: Specific session identifier
• utility_classes: Utility classes provided to students during the exam
• solution: Gold solution created by Prof. Viroli (one of many possible solutions)
• test: JUnit test cases that must be passed

Benchmark Significance

JAB offers several unique advantages for evaluating LLMs on Java programming tasks:

1. Educational Validity: Problems are designed by education experts to test conceptual understanding
2. Comprehensive Assessment: Tasks require full class implementations rather than just completing isolated functions
3. Object-Oriented Focus: Specifically targets OOP concepts including inheritance, polymorphism, encapsulation, and interfaces
4. Longitudinal Coverage: Spans a decade of real academic assessments (2014-2024)
5. Test-Driven Evaluation: Uses rigorous JUnit test suites for objective assessment

License

This dataset is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.

You are free to:

• Share — copy and redistribute the material in any medium or format

Under the following terms:

• Attribution — You must give appropriate credit (see below), provide a link to the license, and indicate if changes were made.
• NonCommercial — You may not use the material for commercial purposes.
• NoDerivatives — If you remix, transform, or build upon the material, you may not distribute the modified material.

For details, see: https://creativecommons.org/licenses/by-nc-nd/4.0/

Dataset Owner: Mirko Viroli
License Notice: © Mirko Viroli — CC BY-NC-ND 4.0

Citation

If you use JAB in your research, please cite:

@article{COCCHIERI2026113033,
title = {Java academic benchmark: Exam-based evaluation of LLMs on object-oriented programming},
journal = {Journal of Systems and Software},
pages = {113033},
year = {2026},
issn = {0164-1212},
doi = {https://doi.org/10.1016/j.jss.2026.113033},
url = {https://www.sciencedirect.com/science/article/pii/S0164121226002669},
author = {Alessio Cocchieri and Luca Ragazzi and Gianluca Aguzzi and Giacomo Frisoni and Lorenzo Molfetta and Gianluca Moro and Mirko Viroli},
keywords = {Large language models, Java, Object-oriented programming, Coding benchmark},
abstract = {Current code generation benchmarks largely overlook object-oriented programming (OOP) skills, leaving open whether large language models (LLMs) can effectively apply OOP principles to structured programming tasks. The dominance and permissiveness of Python in most benchmarks further obscure model weaknesses in core OOP principles. We introduce the Java Academic Benchmark (JAB), based on 103 authentic Java exams collected over a decade at a major university, together with 506 expert-written JUnit tests, to rigorously evaluate advanced Java programming competence with a strong focus on OOP. To complement execution-based scoring, we propose KODE, an LLM-as-a-Judge framework that assesses OOP adherence across four pedagogical dimensions. We evaluate 27 LLMs under two resolution strategies: single-attempt (one-shot) and agentic (iterative refinement with compiler and test feedback). Results reveal that, under our evaluation protocol, larger closed models match or surpass bachelor-level OOP students–especially under agentic resolution–while smaller open models lag behind. JAB’s class-level design enables fine-grained error analysis, exposing recurring misconceptions. Editor’s note: Open Science material was validated by the Journal of Systems and Software Open Science Board.}
}

Contact

For questions or further information about the JAB benchmark, please contact:

• Mirko Viroli