DafnyBench: A Benchmark for Formal Software Verification
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910483542441984 |
|---|---|
| author | Loughridge, Chloe Sun, Qinyi Ahrenbach, Seth Cassano, Federico Sun, Chuyue Sheng, Ying Mudide, Anish Misu, Md Rakib Hossain Amin, Nada Tegmark, Max |
| author_facet | Loughridge, Chloe Sun, Qinyi Ahrenbach, Seth Cassano, Federico Sun, Chuyue Sheng, Ying Mudide, Anish Misu, Md Rakib Hossain Amin, Nada Tegmark, Max |
| contents | We introduce DafnyBench, the largest benchmark of its kind for training and evaluating machine learning systems for formal software verification. We test the ability of LLMs such as GPT-4 and Claude 3 to auto-generate enough hints for the Dafny formal verification engine to successfully verify over 750 programs with about 53,000 lines of code. The best model and prompting scheme achieved 68% success rate, and we quantify how this rate improves when retrying with error message feedback and how it deteriorates with the amount of required code and hints. We hope that DafnyBench will enable rapid improvements from this baseline as LLMs and verification techniques grow in quality. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_08467 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | DafnyBench: A Benchmark for Formal Software Verification Loughridge, Chloe Sun, Qinyi Ahrenbach, Seth Cassano, Federico Sun, Chuyue Sheng, Ying Mudide, Anish Misu, Md Rakib Hossain Amin, Nada Tegmark, Max Software Engineering Artificial Intelligence Machine Learning Programming Languages We introduce DafnyBench, the largest benchmark of its kind for training and evaluating machine learning systems for formal software verification. We test the ability of LLMs such as GPT-4 and Claude 3 to auto-generate enough hints for the Dafny formal verification engine to successfully verify over 750 programs with about 53,000 lines of code. The best model and prompting scheme achieved 68% success rate, and we quantify how this rate improves when retrying with error message feedback and how it deteriorates with the amount of required code and hints. We hope that DafnyBench will enable rapid improvements from this baseline as LLMs and verification techniques grow in quality. |
| title | DafnyBench: A Benchmark for Formal Software Verification |
| topic | Software Engineering Artificial Intelligence Machine Learning Programming Languages |
| url | https://arxiv.org/abs/2406.08467 |