Code2LoRA: Hypernetwork-Generated Adapters for Code Language Models under Software Evolution

论文概要

研究领域: NLP
作者: Liliana Hotsko, Yinxi Li, Yuntian Deng
发布时间: 2026-06-04
arXiv: 2606.06492

中文摘要

代码语言模型需要仓库级上下文来解析导入、API和项目约定。现有方法通过长输入（RAG或依赖分析检索）或逐仓库微调/LoRA来注入知识——在仓库规模上代价高昂，且对演进代码库脆弱。我们提出Code2LoRA，一种超网络框架，生成仓库特定的LoRA适配器，零推理token开销注入仓库知识。Code2LoRA支持两种场景：Code2LoRA-Static将单个仓库快照转换为适配器，适用于稳定代码库理解；Code2LoRA-Evo通过GRU隐状态维护适配器，每代码diff更新，适用于演进代码库主动开发。为评估Code2LoRA，我们构建RepoPeftBench基准：604个Python仓库，静态轨道40K训练/12K测试断言补全任务，演进轨道215K提交训练/87K提交测试任务。静态轨道Code2LoRA-Static达63.8%跨仓库和66.2%仓库内精确匹配，与逐仓库LoRA上界持平；演进轨道Code2LoRA-Evo达60.3%跨仓库精确匹配（+5.2pp优于单一共享LoRA）。

原文摘要

Code language models need repository-level context to resolve imports, APIs, and project conventions. Existing methods inject this knowledge as long inputs (retrieved through RAG or dependency analysis) or through per-repository fine-tuning and LoRA -- costly at repository scale and brittle to evolving codebases. We introduce Code2LoRA, a hypernetwork framework that generates repository-specific LoRA adapters, effectively injecting repository knowledge with zero inference-time token overhead. Code2LoRA supports two usage scenarios: Code2LoRA-Static converts a single repository snapshot into an adapter, suitable for comprehension of stable codebases; while Code2LoRA-Evo maintains an adapter backed by a GRU hidden state updated per code diff, suitable for active development of evolving codeb...

自动采集于 2026-06-07

#论文 #arXiv #NLP #小凯