Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

InfCode-C++: Intent-Guided Semantic Retrieval and AST-Structured Search for C++ Issue Resolution

Qingao Dong, Mengfei Wang, Hengzhi Zhang, Zhichao Li, Yuan Yuan, Mu Li, Xiang Gao, Hailong Sun, Chunming Hu, Weifeng Lv

TL;DR

INFCODE-C++ is introduced, the first C++-aware autonomous system for end-to-end issue resolution and highlights the need for language-aware reasoning in multi-language software agents and establishes a foundation for future research on scalable, LLM-driven repair for complex, statically typed ecosystems.

Abstract

Large language model (LLM) agents have recently shown strong performance on repository-level issue resolution, but existing systems are almost exclusively designed for Python and rely heavily on lexical retrieval and shallow code navigation. These approaches transfer poorly to C++ projects, where overloaded identifiers, nested namespaces, template instantiations, and deep control-flow structures make context retrieval and fault localization substantially more difficult. As a result, state-of-the-art Python-oriented agents show a drastic performance drop on the C++ subset of MultiSWE-bench. We introduce INFCODE-C++, the first C++-aware autonomous system for end-to-end issue resolution. The system combines two complementary retrieval mechanisms -- semantic code-intent retrieval and deterministic AST-structured querying -- to construct accurate, language-aware context for repair.These components enable precise localization and robust patch synthesis in large, statically typed C++ repositories. Evaluated on the \texttt{MultiSWE-bench-CPP} benchmark, INFCODE-C++ achieves a resolution rate of 25.58\%, outperforming the strongest prior agent by 10.85 percentage points and more than doubling the performance of MSWE-agent. Ablation and behavioral studies further demonstrate the critical role of semantic retrieval, structural analysis, and accurate reproduction in C++ issue resolution. INFCODE-C++ highlights the need for language-aware reasoning in multi-language software agents and establishes a foundation for future research on scalable, LLM-driven repair for complex, statically typed ecosystems.

InfCode-C++: Intent-Guided Semantic Retrieval and AST-Structured Search for C++ Issue Resolution

TL;DR

INFCODE-C++ is introduced, the first C++-aware autonomous system for end-to-end issue resolution and highlights the need for language-aware reasoning in multi-language software agents and establishes a foundation for future research on scalable, LLM-driven repair for complex, statically typed ecosystems.

Abstract

Large language model (LLM) agents have recently shown strong performance on repository-level issue resolution, but existing systems are almost exclusively designed for Python and rely heavily on lexical retrieval and shallow code navigation. These approaches transfer poorly to C++ projects, where overloaded identifiers, nested namespaces, template instantiations, and deep control-flow structures make context retrieval and fault localization substantially more difficult. As a result, state-of-the-art Python-oriented agents show a drastic performance drop on the C++ subset of MultiSWE-bench. We introduce INFCODE-C++, the first C++-aware autonomous system for end-to-end issue resolution. The system combines two complementary retrieval mechanisms -- semantic code-intent retrieval and deterministic AST-structured querying -- to construct accurate, language-aware context for repair.These components enable precise localization and robust patch synthesis in large, statically typed C++ repositories. Evaluated on the \texttt{MultiSWE-bench-CPP} benchmark, INFCODE-C++ achieves a resolution rate of 25.58\%, outperforming the strongest prior agent by 10.85 percentage points and more than doubling the performance of MSWE-agent. Ablation and behavioral studies further demonstrate the critical role of semantic retrieval, structural analysis, and accurate reproduction in C++ issue resolution. INFCODE-C++ highlights the need for language-aware reasoning in multi-language software agents and establishes a foundation for future research on scalable, LLM-driven repair for complex, statically typed ecosystems.

Paper Structure

This paper contains 42 sections, 13 equations, 3 figures, 4 tables, 1 algorithm.

Table of Contents

  1. INTRODUCTION
  2. Motivation Example
  3. Lexical retrieval fails in the presence of overloaded identifiers.
  4. Structured retrieval resolves ambiguity.
  5. METHODOLOGY
  6. Problem Definition
  7. System Overview
  8. Repository Parsing
  9. AST-Based Structural Index.
  10. Semantic Code-Intent Index.
  11. Issue Reproduction
  12. Patch Agent: Context Retrieval and Patch Generation
  13. Semantic Retrieval via Code Intent
  14. Structural Retrieval via AST Querying
  15. Patch Selection
  16. ...and 27 more sections

Figures (3)

  • Figure 1: Motivation example illustrating the ambiguity of lexical retrieval and the correctness of AST-structured querying in C++ codebases.
  • Figure 2: Overall system architecture of InfCode-C++. The framework consists of three collaborative agents: (1) the Reproducer Agent, which reconstructs the failure and produces a concrete reproducing test; (2) the Patch Agent, which performs semantic intent retrieval, AST-structured querying, and multi-step patch synthesis using a rich tool kit; (3) the Selector Agent, which prunes, validates, and ranks candidate patches through regression testing, extra test generation, and agent voting.
  • Figure 3: Instance-level overlap of resolved issues on Multi-SWE-bench-CPP across InfCode and representative baselines.