The Year in AI Papers
The year/Independent research

Paper 2509.16198

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, et al.

Published
Sep 2025
Research lab
Independent
Citations
12
GitHub
Not linked

01 In brief

Summary

The paper introduces the Repository Planning Graph (RPG), a structured representation that unifies proposal-level and implementation-level planning for generating complete software repositories from high-level specifications.

RPG encodes capabilities, file structures, data flows, and functions as nodes and edges, replacing ambiguous natural language plans.

The authors develop ZeroRepo, a graph-driven framework that constructs RPG through three stages: proposal-level construction, implementation-level construction, and graph-guided code generation with test validation.

To evaluate, they build RepoCraft, a benchmark of six real-world projects (scikit-learn, pandas, sympy, statsmodels, requests, django) with 1,052 tasks.

On RepoCraft, ZeroRepo achieves 81.5% functional coverage and 69.7% test accuracy, outperforming the strongest baseline (Claude Code) by 27.3 and 35.8 points respectively, while generating repositories on average 3.9× larger (36K LOC, 445K tokens).

Analysis shows RPG captures complex dependencies, enables near-linear scaling of functionality and code size, and improves agent localization efficiency.

The work demonstrates that graph-based representations provide a robust foundation for long-horizon, large-scale repository generation, with data and code available at https://github.com/microsoft/RPG-ZeroRepo.

02 From the paper

Abstract

Large language models excel at generating individual functions or single files of code, yet generating complete repositories from scratch remains a fundamental challenge. This capability is key to building coherent software systems from high-level specifications and realizing the full potential of automated code generation. The process requires planning at two levels: deciding what features and modules to build (proposal stage) and defining their implementation details (implementation stage). Current approaches rely on natural language planning, which often produces unclear specifications, misaligned components, and brittle designs due to its inherent ambiguity and lack of structure. To address these limitations, we introduce the Repository Planning Graph (RPG), a structured representation that encodes capabilities, file structures, data flows, and functions in a unified graph. By replacing free-form natural language with an explicit blueprint, RPG enables consistent long-horizon planning for repository generation. Building on RPG, we develop ZeroRepo, a graph-driven framework that operates in three stages: proposal-level planning, implementation-level construction, and graph-guided code generation with test validation. To evaluate, we construct RepoCraft, a benchmark of six real-world projects with 1,052 tasks. On RepoCraft, ZeroRepo produces nearly 36K Code Lines and 445K Code Tokens, on average 3.9$\times$ larger than the strongest baseline (Claude Code), and 68$\times$ larger than other baselines. It achieves 81.5% coverage and 69.7% test accuracy, improving over Claude Code by 27.3 and 35.8 points. Further analysis shows that RPG models complex dependencies, enables more sophisticated planning through near-linear scaling, and improves agent understanding of repositories, thus accelerating localization. Our data and code are available at https://github.com/microsoft/RPG-ZeroRepo.