Research map / Large language models / Agents, code and math
RPG: A Repository Planning Graph for Unified and Scalable Codebase Generation
Jane Luo, Xin Zhang, Steven Liu, Jie Wu, Jianfeng Liu, Yiming Huang, Yangyu Huang, Chengyu Yin, Ying Xin, Yuefeng Zhan, Hao Sun, Qi Chen and 2 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 |
|---|---|---|---|
| use RPG-structured planning instead of unstructured planning (both without EFT) (ablation) | ZeroRepo · RepoCraft MLKit-Py | coverage (%): 59.6 -> 87.2 | Improved |
| use RPG-structured planning instead of unstructured planning (both without EFT) (ablation) | ZeroRepo · RepoCraft MLKit-Py | lines of code: 25,202 LOC; 7x volume increase | Improved |
| add the EpiCoder Feature Tree knowledge base (ablation) | ZeroRepo (o3-mini) · RepoCraft MLKit-Py | coverage (%) speed: 97.9% by Iteration 40 without EFT matches EFT version at Iteration 30 | Improved |
| add the EpiCoder Feature Tree to LOC growth (ablation) | ZeroRepo (o3-mini) · RepoCraft MLKit-Py | LOC vs iteration fit: y ≈ 983x + 2992 vs y ≈ 800x + 989 | Improved |
| replace Claude Code with ZeroRepo (o3-mini) as agent | Claude Code CLI · RepoCraft | coverage (%): 54.2 -> 81.5 | Improved |
| replace Claude Code with ZeroRepo (o3-mini) as agent | Claude Code CLI · RepoCraft | pass rate (%): 33.9 -> 69.7 | Improved |
| use Qwen3-Coder backbone instead of o3-mini for ZeroRepo | ZeroRepo · RepoCraft | coverage / pass rate / LOC: 81.5 -> 75.1 coverage; 69.7 -> 57.3 pass rate; 23,977.3 -> 36,941.0 LOC | Mixed |
| increase planning iterations from 5 to 30 | ZeroRepo · RepoCraft | coverage (%): 70.2 -> nearly 96 | Improved |
| add graph-guided search (RPG) to agent localization (ablation) | coding agent · Integration testing, incremental development, debugging | localization steps: reduces effort by 30-50% | Improved |
| use o3-mini as automated evaluator versus human | ZeroRepo evaluation · RepoCraft | Pearson correlation with human: 0.893 coverage and 0.958 novelty | Improved |
Closest papers by meaning
- NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents (ICML 2026 · similarity 0.89)
- RepoGraph: Enhancing AI Software Engineering with Repository-level Code Graph (ICLR 2025 · similarity 0.88)
- Code Graph Model (CGM): A Graph-Integrated Large Language Model for Repository-Level Software Engineering Tasks (NeurIPS 2025 · similarity 0.87)
- RepoMaster: Autonomous Exploration and Understanding of GitHub Repositories for Complex Task Solving (NeurIPS 2025 · similarity 0.87)
- GitTaskBench: A Benchmark for Code Agents Solving Real-World Tasks Through Code Repository Leveraging (AAAI 2026 · similarity 0.86)
- Build-and-Find: An Effort-Aware Protocol for Evaluating Agent-Managed Codebases (NeurIPS 2026 · similarity 0.86)
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.
