Saved in:
Bibliographic Details
Main Authors: Yang, Jian, Guo, Shawn, Jing, Lin, Zhang, Wei, Liu, Aishan, Hao, Chuan, Li, Zhoujun, Zhao, Wayne Xin, Liu, Xianglong, Lv, Weifeng, Dai, Bryan
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.13472
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912765371744256
author Yang, Jian
Guo, Shawn
Jing, Lin
Zhang, Wei
Liu, Aishan
Hao, Chuan
Li, Zhoujun
Zhao, Wayne Xin
Liu, Xianglong
Lv, Weifeng
Dai, Bryan
author_facet Yang, Jian
Guo, Shawn
Jing, Lin
Zhang, Wei
Liu, Aishan
Hao, Chuan
Li, Zhoujun
Zhao, Wayne Xin
Liu, Xianglong
Lv, Weifeng
Dai, Bryan
contents Code large language models (Code LLMs) are powerful but costly to train, with scaling laws predicting performance from model size, data, and compute. However, different programming languages (PLs) have varying impacts during pre-training that significantly affect base model performance, leading to inaccurate performance prediction. Besides, existing works focus on language-agnostic settings, neglecting the inherently multilingual nature of modern software development. Therefore, it is first necessary to investigate the scaling laws of different PLs, and then consider their mutual influences to arrive at the final multilingual scaling law. In this paper, we present the first systematic exploration of scaling laws for multilingual code pre-training, conducting over 1000+ experiments (Equivalent to 336,000+ H800 hours) across multiple PLs, model sizes (0.2B to 14B parameters), and dataset sizes (1T tokens). We establish comprehensive scaling laws for code LLMs across multiple PLs, revealing that interpreted languages (e.g., Python) benefit more from increased model size and data than compiled languages (e.g., Rust). The study demonstrates that multilingual pre-training provides synergistic benefits, particularly between syntactically similar PLs. Further, the pre-training strategy of the parallel pairing (concatenating code snippets with their translations) significantly enhances cross-lingual abilities with favorable scaling properties. Finally, a proportion-dependent multilingual scaling law is proposed to optimally allocate training tokens by prioritizing high-utility PLs (e.g., Python), balancing high-synergy pairs (e.g., JavaScript-TypeScript), and reducing allocation to fast-saturating languages (Rust), achieving superior average performance across all PLs compared to uniform distribution under the same compute budget.
format Preprint
id arxiv_https___arxiv_org_abs_2512_13472
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Scaling Laws for Code: Every Programming Language Matters
Yang, Jian
Guo, Shawn
Jing, Lin
Zhang, Wei
Liu, Aishan
Hao, Chuan
Li, Zhoujun
Zhao, Wayne Xin
Liu, Xianglong
Lv, Weifeng
Dai, Bryan
Computation and Language
Code large language models (Code LLMs) are powerful but costly to train, with scaling laws predicting performance from model size, data, and compute. However, different programming languages (PLs) have varying impacts during pre-training that significantly affect base model performance, leading to inaccurate performance prediction. Besides, existing works focus on language-agnostic settings, neglecting the inherently multilingual nature of modern software development. Therefore, it is first necessary to investigate the scaling laws of different PLs, and then consider their mutual influences to arrive at the final multilingual scaling law. In this paper, we present the first systematic exploration of scaling laws for multilingual code pre-training, conducting over 1000+ experiments (Equivalent to 336,000+ H800 hours) across multiple PLs, model sizes (0.2B to 14B parameters), and dataset sizes (1T tokens). We establish comprehensive scaling laws for code LLMs across multiple PLs, revealing that interpreted languages (e.g., Python) benefit more from increased model size and data than compiled languages (e.g., Rust). The study demonstrates that multilingual pre-training provides synergistic benefits, particularly between syntactically similar PLs. Further, the pre-training strategy of the parallel pairing (concatenating code snippets with their translations) significantly enhances cross-lingual abilities with favorable scaling properties. Finally, a proportion-dependent multilingual scaling law is proposed to optimally allocate training tokens by prioritizing high-utility PLs (e.g., Python), balancing high-synergy pairs (e.g., JavaScript-TypeScript), and reducing allocation to fast-saturating languages (Rust), achieving superior average performance across all PLs compared to uniform distribution under the same compute budget.
title Scaling Laws for Code: Every Programming Language Matters
topic Computation and Language
url https://arxiv.org/abs/2512.13472