Research map / Vision-language and multimodal / Human and embodied modeling
Closing the Loop: Universal Repository Representation with RPG-Encoder
Jane Luo, Chengyu Yin, Xin Zhang, Qingtao Li, Steven Liu, Yiming Huang, Jie Wu, Hao Liu, Yangyu Huang, Yu Kang, Fangkai Yang, Ying Xin and 1 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add semantic features to the graph (ablation) | RPG-Encoder (GPT-4o) · SWE-bench Live function-level | Acc@1: 43.1 -> 50.5 | Improved |
| add dependency edges to the graph (ablation) | RPG-Encoder (GPT-4o) · SWE-bench Live file-level | Acc@1: 58.4 -> 69.2 | Improved |
| add semantic features to the graph (ablation) | RPG-Encoder (GPT-4.1) · SWE-bench Live function-level | Acc@1: 57.4 -> 64.7 | Improved |
| add dependency edges to the graph (ablation) | RPG-Encoder (GPT-4.1) · SWE-bench Live function-level | Acc@1: 63.7 -> 64.7 | No clear effect |
| add function metadata to nodes (ablation) | ZeroRepo-RPG (GPT-5-mini) · scikit-learn reconstruction | Pass rate: 74.1 -> 82.8 | Improved |
| add file and function metadata to nodes (ablation) | ZeroRepo-RPG (GPT-5-mini) · scikit-learn reconstruction | #Files: 157 -> 256 | Improved |
| add all node metadata (ablation) | ZeroRepo-RPG (GPT-5-mini) · scikit-learn reconstruction | Coverage: 87.2 -> 100.0 | Improved |
| replace API documentation with RPG as reconstruction source | ZeroRepo (GPT-5-mini) · RepoCraft | Coverage: 74.2 -> 98.5 | Improved |
| replace API documentation with RPG as reconstruction source | ZeroRepo (GPT-4.1) · RepoCraft | Coverage: 64.6 -> 93.5 | Improved |
| replace baseline localizers with RPG-Encoder | Claude-4.5-Sonnet · SWE-bench Verified function-level | Acc@5: 79.3 -> 93.7 | Improved |
| replace baseline localizers with RPG-Encoder | GPT-5 · SWE-bench Live function-level | Acc@5: 76.2 -> 87.8 | Improved |
| use RPG instead of baseline navigation for efficiency | GPT-4.1 · SWE-bench Verified | Acc@5/Cost: 3.10 -> 4.63 | Improved |
Closest papers by meaning
- RepoMirage: Probing Repository Context Reasoning in Code Agents with Perturbations (NeurIPS 2026 · similarity 0.88)
- Improving Code Localization with Repository Memory (ICLR 2026 · similarity 0.87)
- NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents (ICML 2026 · similarity 0.87)
- Build-and-Find: An Effort-Aware Protocol for Evaluating Agent-Managed Codebases (NeurIPS 2026 · similarity 0.86)
- RepoGraph: Enhancing AI Software Engineering with Repository-level Code Graph (ICLR 2025 · similarity 0.86)
- BootstrapAgent: Turning Repository Setup into Reusable Agent Knowledge (NeurIPS 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
