SWE-Bench 5G: Benchmarking AI Coding Agents on Telecom Network Engineering Tasks

2026-04-29

AI Summary

This study addresses the lack of domain-specific evaluation benchmarks for AI coding agents in telecommunications by constructing the first defect-repair benchmark tailored to 5G core networks. The benchmark integrates real-world bugs from three open-source projects, packaged within Docker environments equipped with automated testing infrastructure, and introduces a dual testing strategy designed to handle complex software dependencies. To enable controlled assessment of domain knowledge impact, the work innovatively incorporates a 3GPP specification context injection mechanism. Experimental results demonstrate that while leading large language models achieve over 91% accuracy in bug diagnosis, their repair success rates remain low at 10%–30%. Injecting 3GPP specifications significantly improves repair performance for specification-related defects, though it yields limited gains for general defensive checks.

Abstract

AI coding agents demonstrate strong performance on general-purpose software benchmarks. However, their ability to handle 5G network engineering tasks remains unexplored. We propose SWE-Bench~5G, the first benchmark designed to investigate whether AI coding agents can resolve real-world bugs in 5G core network software. The benchmark collects task instances from three open-source 5G projects, packages each as a self-contained Docker environment with automated fail-to-pass tests, and provides a dual test strategy tailored to the complex runtime dependencies of telecom code. In addition, for instances whose original issues reference 3GPP specification clauses, we construct concise specification context documents, enabling controlled evaluation of whether domain knowledge improves agent performance. Experiments on four LLMs reveal that all models diagnose bugs at rates exceeding 91\%, yet resolve rates remain between 10\% and 30\%, suggesting that both iterative code editing capability and domain knowledge play important roles. The specification injection experiment further confirms that 3GPP excerpts improve resolve rates on specification-dependent bugs, while the gains on generic defensive checks remain limited, indicating that the effect of domain knowledge is conditional on bug type.

Problem

• AI coding agents
• 5G network engineering
• software bugs
• telecom software
• domain knowledge

Innovation

• SWE-Bench 5G
• AI coding agents
• 5G core network
• 3GPP specification
• domain knowledge injection