Multilingual Code Generation

Follow this topic Ask about this topic

Multilingual code generation studies how models turn instructions into working programs across programming languages and human languages, supporting polyglot software and broader access to coding tools. Common approaches include training on multilingual code and prompts, translating tasks across languages, using language-conditioned adapters, and improving models with execution-based feedback or reinforcement learning. Evaluation uses executable benchmarks spanning many programming languages, alongside measures of instruction following, code quality, and agent behavior. Performance remains uneven across languages: success in Python does not reliably transfer, and prompt-language effects vary, while compile errors and contamination make fair comparisons difficult.

Latest papers

16 papers1 paper in the last four weeks, with none the four weeks before. 0.0% of all new papers.

CodeForge-MA: Execution-Verified Multi-Agent Learning with Language-Conditioned LoRA for Multilingual Code Generation
Oct 4, 2026Zhizhou Gu, Xianting Wu, Siyu Gu +2RL for Code Generation Synthetic Data Generation for Language Models
Northeastern University Bellevue, USA · Columbia University New York, USA · University of Pennsylvania Philadelphia, USA +2
Terminal-Bench-LILT: Multilingual Agentic Coding Benchmark Grounded in Language, Region, and Culture
Aug 11, 2026Yunsu Kim, Kaden Uhlig, Ashwin Purohit +10AI Agent Benchmarks LLM Agent Evaluation
LILT, Inc.
The Best Programming Language for Tokenmaxxing: An Investigation of Coding Agent Behavior Across Programming Languages
Jul 24, 2026Zixuan Wu, Carolyn Jane Anderson, Arjun GuhaAI Coding Agents Code Generation
Northeastern University · Wellesley College
Large Language Models for Code Generation from Multilingual Prompts: A Curated Benchmark and a Study on Code Quality
Jul 16, 2026Saima Afrin, Alessandro Midolo, Camilo Escobar-Velásquez +5Multilingual Language Model Evaluation Code Generation
William & Mary, VA, USA · University of Catania, Italy · Universidad de los Andes, Colombia +2
Multi-LCB: Extending LiveCodeBench to Multiple Programming Languages
Jun 18, 2026Maria Ivanova, Pavel Zadorozhny, Rodion Levichev +5Code Generation Code Generation Evaluation
Yandex School of Data Analysis, Applied AI Institute · GigaCode
Beyond Pass Rate: A Multilingual, Execution-Grounded Evaluation of Open Code LLMs
Jun 7, 2026Sayed Erfan ArefinCode Generation Evaluation Code Language Models
Department of Computer Science University of Dayton Dayton, OH 45469, USA
CodeGolf Bench: A Multi-Language Benchmark for Evaluating Concise Code Generation Capabilities of Large Language Models
May 28, 2026Vedant PadwalLanguage Model Generation Evaluation Code Generation
Independent
Delulu: A Verified Multi-Lingual Benchmark for Code Hallucination Detection in Fill-in-the-Middle Tasks
May 7, 2026Mahdi Erfanian, Nelson Daniel Troncoso, Aashna Garg +4Language Model Error Detection Language Model Generation Evaluation
University of Illinois Chicago · Microsoft
Large Language Models for Multilingual Code Intelligence: A Survey
Apr 27, 2026Chao Jiang, Dugang Liu, Cheng Wen +6Code Translation Code Generation
College of Computer Science and Software Engineering, Shenzhen University, China · Guangzhou Institute of Technology, Xidian University, China · Guangzhou University of Software, China +2
Aligned Multi-View Scripts for Universal Chart-to-Code Generation
Apr 27, 2026Zhihan Zhang, Lizi LiaoCode Generation Parameter-Efficient Fine-Tuning
School of Computing and Information Systems, Singapore Management University
FLeX: Fourier-based Low-rank EXpansion for multilingual transfer
Apr 6, 2026Gaurav NarasimhanLLM Fine-Tuning Multilingual Code Generation
Department of Computer Science, Stanford University
Chinese Language Is Not More Efficient Than English in Vibe Coding: A Preliminary Study on Token Cost and Problem-Solving Rate
Apr 6, 2026Simiao Ren, Xingyu Shen, Yuchen Zhou +3LLM Inference Efficiency Multilingual Code Generation
Tsang · ∗,†Scam.ai
IndustryCode: A Benchmark for Industry Code Generation
Apr 3, 2026Puyu Zeng, Zhaoxi Wang, Zhixu Duan +7Code Generation Code Generation Benchmarks
Shanghai Jiao Tong University, Shanghai, China · Alibaba Group, Hangzhou, China · Tsinghua University, Beijing, China
Agnostics: Learning to Code in Any Programming Language via Reinforcement with a Universal Learning Environment
Aug 6, 2025Aleksander Boruch-Gruszecki, Yangtian Zi, Zixuan Wu +4RL for Code Generation Code Generation
1Northeastern University · University of Texas · 3Wellesley College
IFEvalCode: Controlled Code Generation
Jul 30, 2025Jian Yang, Wei Zhang, Shukai Liu +9Code Generation Code Language Models
Beihang University · M-A-P · OPPO
EffiBench-X: A Multi-Language Benchmark for Measuring Efficiency of LLM-Generated Code
May 19, 2025Yuhao Qing, Boyu Zhu, Mingzhe Du +9Code Generation Benchmarks LLM Code Generation
HKU · UCL · NTU +6

Oct 4, 2026J/K move · Enter open · S save

CodeForge-MA: Execution-Verified Multi-Agent Learning with Language-Conditioned LoRA for Multilingual Code Generation

Zhizhou Gu, Xianting Wu, Siyu Gu, Tian Zhang+1

Northeastern University Bellevue, USA · Columbia University New York, USA · University of Pennsylvania Philadelphia, USA+2

RL for Code Generation Synthetic Data Generation for Language Models Multi-Agent Coding Multilingual Code Generation Parameter-Efficient Fine-Tuning

Open paper AskPDF arXiv

Large language models for code generation often fail on execution, multilingual coverage, and contamination control, especially under frozen backbone constraints. We present CodeForge-MA, a unified framework that improves code synthesis through a multi-agent data forge, execution verified reinforced instruction tuning, and a language conditioned mixture of LoRA adapters. Four specialized agents, Composer, Reviewer, Executor, and Curator, iteratively refine instruction code pairs, validate them with tests, and filter duplicates and benchmark leakage. During training, we combine masked supervised fine tuning with a test driven reinforcement objective to align generations with executable correctness. For the larger model, we use sparse expert routing over low rank adapters to improve cross language transfer while keeping the base model unchanged at inference. Experiments show that joint data, objective, and adapter design yields robust gains across programming languages.