Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2510.04605 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915534017134592 |
|---|---|
| author | Zhang, Jingyao Li, Tianlin Zhang, Xiaoyu Hu, Qiang Shi, Bin |
| author_facet | Zhang, Jingyao Li, Tianlin Zhang, Xiaoyu Hu, Qiang Shi, Bin |
| contents | Autoregressive Large Language Models (AR-LLMs) are widely used in software engineering (SE) but face limitations in processing code structure information and suffer from high inference latency. Diffusion LLMs (DLLMs) offer a promising alternative with global bidirectional encoding and decoupled generation steps. This work presents the first comprehensive evaluation of DLLMs across the software development lifecycle, including code generation, defect detection, and program repair. On a large-scale benchmark of 52,937 tasks, 7Bparameter DLLMs outperform AR-LLMs with a 30% average accuracy improvement achieving a 113% gain on cross-file repair, while maintaining superior efficiency and reduced latency. Our results establish DLLMs as a superior paradigm for SE tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_04605 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Exploring the Power of Diffusion Large Language Models for Software Engineering: An Empirical Investigation Zhang, Jingyao Li, Tianlin Zhang, Xiaoyu Hu, Qiang Shi, Bin Software Engineering Autoregressive Large Language Models (AR-LLMs) are widely used in software engineering (SE) but face limitations in processing code structure information and suffer from high inference latency. Diffusion LLMs (DLLMs) offer a promising alternative with global bidirectional encoding and decoupled generation steps. This work presents the first comprehensive evaluation of DLLMs across the software development lifecycle, including code generation, defect detection, and program repair. On a large-scale benchmark of 52,937 tasks, 7Bparameter DLLMs outperform AR-LLMs with a 30% average accuracy improvement achieving a 113% gain on cross-file repair, while maintaining superior efficiency and reduced latency. Our results establish DLLMs as a superior paradigm for SE tasks. |
| title | Exploring the Power of Diffusion Large Language Models for Software Engineering: An Empirical Investigation |
| topic | Software Engineering |
| url | https://arxiv.org/abs/2510.04605 |