Large Language Models for Multilingual Code Intelligence: A Survey

arXiv cs.LG / 4/30/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

The survey examines how large language models are being used for AI-assisted software engineering, noting performance gaps across programming languages.
It highlights that existing research is often biased toward high-resource languages like Python, while languages such as Rust and OCaml still lag behind.
The work focuses on two core multilingual code intelligence tasks: generating code in multiple languages from shared natural-language requirements and translating code across languages while preserving semantics.
It reviews representative approaches, benchmarks, and evaluation metrics, and discusses key challenges and opportunities for reliable cross-language generalization.
The survey frames multilingual, trustworthy code intelligence as essential because real-world software systems are typically “polyglot.”

Abstract

Large language models have transformed AI-assisted software engineering, but current research remains biased toward high-resource languages such as Python, with weaker performance in languages like Rust and OCaml. Since real-world systems are inherently polyglot, robust multilingual code intelligence is crucial. This survey focuses on two key tasks: multilingual code generation from shared natural-language requirements, and multilingual code translation that preserves semantics across languages. It reviews representative methods, benchmarks, and evaluation metrics, and highlights challenges and opportunities for trustworthy cross-language generalization.