arXiv cs.AI· Xiaolong Xu, Hao Jiang, Bo Jiang, Zhiwen Deng, Ningyuan Sun, Bingzhou Chen, Jin Chen, Chenxing Wei·· 3 天前AI 评分31
Diff Diffusion：用扩散模型预测编辑路径实现高效代码编辑
Diff Diffusion: Predicting Edit Paths with Diffusion Models for Efficient Code Editing
AI 导读
Diff Diffusion 是一个受扩散模型启发的 LLM 框架，将去噪过程表示为结构化编辑操作，从原始代码片段预测到修改后形式的编辑路径，从而把计算集中在改动区域。
正文 · 原文
Abstract:Diffusion large language models (dLLMs) are appealing for code generation because they can update many tokens in parallel, yet many are optimized for sequence synthesis rather than edit prediction. This setting is poorly matched to modern AI-assisted IDEs, where edit prediction must infer the next code change from recent edits, the current file, and cursor-local context. Such changes are usually sparse and preserve most source tokens, so regenerating a full code segment wastes computation. We propose Diff Diffusion, a diffusion-inspired LLM framework for code editing that predicts edit paths from an original snippet to its revised form. Diff Diffusion represents denoising as structured edit operations, and combines this formulation with multi-stage training and efficient parallel decoding. By focusing computation on changed regions, Diff Diffusion supports accurate low-latency edits and multi-position parallel modification. Across public edit-completion benchmarks, edit-prediction evaluations, and online validation, Diff Diffusion matches or leads autoregressive baselines while delivering 2.78 times faster inference on public edit-completion benchmarks and a 39.0% P50 model-latency reduction in production. This work provides a new modeling perspective for AI applications that require high accuracy and low latency.
| Comments: | 15 pages, 3 figures, 6 tables. Accepted at the NeurIPS 2026 Workshop on Diffusion Language Models: Foundations, Efficiency, and Reasoning |
| Subjects: | Software Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.17831 [cs.SE] |
| (or arXiv:2610.17831v1 [cs.SE] for this version) | |
| https://doi.org/10.48550/arXiv.2610.17831 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Hao Jiang [view email]
[v1]
Tue, 13 Oct 2026 06:49:52 UTC (1,387 KB)
来源：arXiv cs.AI · arxiv.org