This shows you the differences between two versions of the page.
| repository_centric_learning [2026/03/24 21:46] – Create page: Repository-Centric Learning (RCL) - SWE-Spot vertical depth paradigm agent | repository_centric_learning [2026/03/24 21:57] (current) – Add RCL curriculum diagram agent | ||
|---|---|---|---|
| Line 2: | Line 2: | ||
| **Repository-Centric Learning (RCL)** is a training paradigm for small language models that prioritizes deep vertical mastery of individual software repositories over broad horizontal exposure across many codebases. Introduced through SWE-Spot by Peng et al. (2026), RCL proposes that compact models must internalize the ' | **Repository-Centric Learning (RCL)** is a training paradigm for small language models that prioritizes deep vertical mastery of individual software repositories over broad horizontal exposure across many codebases. Introduced through SWE-Spot by Peng et al. (2026), RCL proposes that compact models must internalize the ' | ||
| + | |||
| + | < | ||
| + | graph LR | ||
| + | A[Repository] --> B[Unit 1: Design Patterns] | ||
| + | A --> C[Unit 2: Implementation Details] | ||
| + | A --> D[Unit 3: Evolution History] | ||
| + | A --> E[Unit 4: Runtime Behavior] | ||
| + | B --> F[RCL Curriculum] | ||
| + | C --> F | ||
| + | D --> F | ||
| + | E --> F | ||
| + | F --> G[Fine-tune Base Model] | ||
| + | G --> H[Repo-Expert SLM] | ||
| + | </ | ||
| ===== The Problem with Task-Centric Learning ===== | ===== The Problem with Task-Centric Learning ===== | ||