CodeR: Issue Resolving with Multi-Agent and Task Graphs

2024/06/03 by Dong Chen, Chen, Dong, Shaoxin Lin +31 · 20 citations
Computer Science · #Advanced Graph Neural Networks #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #FOS: Computer and information sciences #Software Engineering (cs.SE) #Topic Modeling #Web Data Mining and Analysis

paper · pdf · doi:10.48550/arxiv.2406.01304

openalex publication_date 2024/06/03 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

GitHub issue resolving recently has attracted significant attention from academia and industry. SWE-bench is proposed to measure the performance in resolving issues. In this paper, we propose CodeR, which adopts a multi-agent framework and pre-defined task graphs to Repair & Resolve reported bugs and add new features within code Repository. On SWE-bench lite, CodeR is able to solve 28.33% of issues, when submitting only once for each issue. We examine the performance impact of each design of CodeR and offer insights to advance this research direction.

Cited by

• Agentic Software Issue Resolution with Large Language Models: A Survey
• A Retrieval-Augmented Generation Approach to Extracting Algorithmic Logic from Neural Networks
• Process-Centric Analysis of Agentic Software Systems
• An Empirical Study of Agent Developer Practices in AI Agent Frameworks
• Is the Cure Still Worse Than the Disease? Test Overfitting by LLMs in Automated Program Repair
• InfCode: Adversarial Iterative Refinement of Tests and Patches for Reliable Software Issue Resolution
• Designing LLM-based Multi-Agent Systems for Software Engineering Tasks: Quality Attributes, Design Patterns and Rationale
• Resolving Java Code Repository Issues with iSWE Agent
• Code Digital Twin: Empowering LLMs with Tacit Knowledge for Complex Software Development
• TDFlow: Agentic Workflows for Test Driven Software Engineering
• TREAT: A Code LLMs Trustworthiness / Reliability Evaluation and Testing Framework
• Helmsman: Autonomous Synthesis of Federated Learning Systems via Collaborative LLM Agents
• PaVeRL-SQL: Text-to-SQL via Partial-Match Rewards and Verbal Reinforcement Learning
• Bridging Solidity Evolution Gaps: An LLM-Enhanced Approach for Smart Contract Compilation Error Resolution
• Heterogeneous Prompting and Execution Feedback for SWE Issue Test Generation and Selection
• Tool-integrated Reinforcement Learning for Repo Deep Search
• SWE-Exp: Experience-Driven Software Issue Resolution
• SWE-Debate: Competitive Multi-Agent Debate for Software Issue Resolution
• IFEvalCode: Controlled Code Generation
• Git Context Controller: Manage the Context of LLM-based Agents like Git

Related

• Beyond Final Code: A Process-Oriented Error Analysis of Software Development Agents in Real-World GitHub Scenarios
• Automated Benchmark Generation for Repository-Level Coding Tasks
• MAGIS: LLM-Based Multi-Agent Framework for GitHub Issue Resolution
• Evaluating Software Development Agents: Patch Patterns, Code Quality, and Issue Complexity in Real-World GitHub Scenarios
• SWE-Mirror: Scaling Issue-Resolving Datasets by Mirroring Issues Across Repositories