研究人员开发了一种名为 TypeProbe 的方法，用于研究预训练代码模型内部如何表示类型信息。通过使用并行的 Java 和 Python 数据集分析这些模型的残差流，研究发现即使在无类型的代码中，跨语言类型表示也会出现。探针还可以线性编码参数和结果类型，证明了对语言和语法变化的鲁棒性。 AI
排序理由 该集群包含一篇详细介绍新研究方法和发现的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
TypeProbe 方法揭示代码模型中的隐藏类型表示
报道来源 [2]
-
TypeProbe: Recovering Type Representations from Hidden States of Pre-trained Code Models
arXiv:2607.08339v1 Announce Type: cross Abstract: State-of-the-art code models achieve impressive performance, yet the extent to which they internally encode type information remains poorly understood. We probe the residual streams of pretrained code models for internal type repr…
-
TypeProbe: Recovering Type Representations from Hidden States of Pre-trained Code Models
State-of-the-art code models achieve impressive performance, yet the extent to which they internally encode type information remains poorly understood. We probe the residual streams of pretrained code models for internal type representations using a parallel dataset of Java and P…