GXPO: Group Cross-Lingual Relative Policy Optimization for Code Generation

Linzheng Chai, Jian Yang, Jiajun Wu, Ensheng Shi, Xianglong Liu
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:12424-12434, 2026.

Abstract

Current reinforcement learning (RL) methods for code generation are predominantly optimized on Python, showing weak generalization to other programming languages (PLs). Although leveraging multilingual solutions offers richer semantics and a wider search landscape, naive independent training across languages suffers from optimization imbalance and fails to effectively transfer knowledge from high-resource languages. We propose Group Cross-lingual Relative Policy Optimization (GXPO), which forms training groups by generating solutions for the same problem in multiple PLs and jointly optimizes language-specific and cross-language signals, enabling more balanced optimization and improved transfer to low-resource PLs. We additionally introduce Multilingual LiveCodeBench (ML-LCB), extending LiveCodeBench to a unified multilingual evaluation setting. On ML-LCB across 8 PLs, GXPO consistently improves performance, with pronounced gains on low-resource PLs, demonstrating scalable multilingual RL for language-consistent code generation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-chai26c, title = {{GXPO}: Group Cross-Lingual Relative Policy Optimization for Code Generation}, author = {Chai, Linzheng and Yang, Jian and Wu, Jiajun and Shi, Ensheng and Liu, Xianglong}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {12424--12434}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/chai26c/chai26c.pdf}, url = {https://proceedings.mlr.press/v306/chai26c.html}, abstract = {Current reinforcement learning (RL) methods for code generation are predominantly optimized on Python, showing weak generalization to other programming languages (PLs). Although leveraging multilingual solutions offers richer semantics and a wider search landscape, naive independent training across languages suffers from optimization imbalance and fails to effectively transfer knowledge from high-resource languages. We propose Group Cross-lingual Relative Policy Optimization (GXPO), which forms training groups by generating solutions for the same problem in multiple PLs and jointly optimizes language-specific and cross-language signals, enabling more balanced optimization and improved transfer to low-resource PLs. We additionally introduce Multilingual LiveCodeBench (ML-LCB), extending LiveCodeBench to a unified multilingual evaluation setting. On ML-LCB across 8 PLs, GXPO consistently improves performance, with pronounced gains on low-resource PLs, demonstrating scalable multilingual RL for language-consistent code generation.} }
Endnote
%0 Conference Paper %T GXPO: Group Cross-Lingual Relative Policy Optimization for Code Generation %A Linzheng Chai %A Jian Yang %A Jiajun Wu %A Ensheng Shi %A Xianglong Liu %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-chai26c %I PMLR %P 12424--12434 %U https://proceedings.mlr.press/v306/chai26c.html %V 306 %X Current reinforcement learning (RL) methods for code generation are predominantly optimized on Python, showing weak generalization to other programming languages (PLs). Although leveraging multilingual solutions offers richer semantics and a wider search landscape, naive independent training across languages suffers from optimization imbalance and fails to effectively transfer knowledge from high-resource languages. We propose Group Cross-lingual Relative Policy Optimization (GXPO), which forms training groups by generating solutions for the same problem in multiple PLs and jointly optimizes language-specific and cross-language signals, enabling more balanced optimization and improved transfer to low-resource PLs. We additionally introduce Multilingual LiveCodeBench (ML-LCB), extending LiveCodeBench to a unified multilingual evaluation setting. On ML-LCB across 8 PLs, GXPO consistently improves performance, with pronounced gains on low-resource PLs, demonstrating scalable multilingual RL for language-consistent code generation.
APA
Chai, L., Yang, J., Wu, J., Shi, E. & Liu, X.. (2026). GXPO: Group Cross-Lingual Relative Policy Optimization for Code Generation. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:12424-12434 Available from https://proceedings.mlr.press/v306/chai26c.html.

Related Material