arXiv cs.CL· Lize Shao, Zichen Xie, Zhizhen Chen, Zhe Zeng, Wenxi Wang·· 12 小时前AI 评分38
SDDC：面向代码生成的自分解扩散框架
Self-Decomposing Diffusion for Code Generation
AI 导读
研究者提出 Self-Decomposing Diffusion for Code（SDDC），先让扩散模型生成描述程序分解与单元间依赖的契约，再据此并行实现各单元。在五个基准和两个 7B 扩散骨干上，SDDC 将 pass@1 最高提升 31.1 个百分点，解码速度比基座模型快 7.5–13.7 倍，并以更低延迟超过微调基线最多 23.5 个点。
正文 · 原文
Abstract:Diffusion language models (DLMs) are emerging as a promising paradigm for code generation, offering the potential for fast, parallel decoding. However, generating complex programs remains challenging, as their interdependent components must maintain consistent interfaces, data flows, and shared assumptions. Aggressive parallel decoding can compromise these dependencies and undermine program correctness, while conservative decoding sacrifices efficiency. To address this trade-off, we propose planning program decomposition before implementation to coordinate interdependent units during parallel generation. We introduce Self-Decomposing Diffusion for Code (SDDC), a novel framework that decomposes complex programs and implements their units in parallel, entirely through diffusion. SDDC first generates a contract specifying the program's decomposition and inter-unit dependencies, then uses it to guide coordinated parallel implementation. Across five benchmarks and two 7B diffusion backbones, SDDC substantially improves the accuracy--latency trade-off, raising pass@1 by up to 31.1 percentage points while decoding 7.5--13.7$\times$ faster than the base models, and outperforming a fine-tuning baseline by up to 23.5 points with lower latency.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.27600 [cs.CL] |
| (or arXiv:2610.27600v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.27600 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Lize Shao [view email]
[v1]
Tue, 20 Oct 2026 04:07:10 UTC (153 KB)
来源：arXiv cs.CL · arxiv.org