Exploring Multi-Lingual Bias of Large Code Models in Code Generation

TL;DR

This paper investigates the multilingual bias in large code models, revealing significant performance disparities across languages and programming languages, and introduces a benchmark for systematic evaluation.

Contribution

The study constructs the first multilingual evaluation benchmark for code generation and provides large-scale analysis of bias in nine popular large code models.

Findings

Chinese instructions reduce performance by at least 13%.

Performance gap between Python and C++ reaches 20.9%.

Multilingual bias is prominent in current large code models.

Abstract

Code generation aims to synthesize code and fulfill functional requirements based on natural language (NL) specifications, which can greatly improve development efficiency. In the era of large language models (LLMs), large code models (LCMs) have been recently proposed to generate source code. LCMs can generate highly feasible solutions for programming problems described in natural language. Despite the effectiveness, we observe a noticeable multilingual bias in the generation performance of LCMs. Specifically, LCMs demonstrate proficiency in generating solutions when provided with instructions in English, yet may falter when faced with semantically equivalent instructions in other NLs such as Chinese. Moreover, the ability of LCMs to generate code exhibits variety across different programming languages (PLs), such as Python and C++. The observed phenomenon indicates the presence of multi-lingual bias within the generative capabilities of LCMs, which has remained unexplored. In this paper, we aim to investigate the multi-lingual bias that exists in current LCMs. First, we initiate our investigation by constructing the first multi-lingual evaluation benchmark X-HumanEval-X, enabling us to systematically evaluate the extent of multi-lingual bias that exists in current LCMs. In our large-scale experiments on nine popular LCMs, we observe a pronounced multi-lingual bias of LCMs in code generation, including multi-NL and multi-PL bias. Specifically, when using Chinese instructions, the code generation capabilities of LCMs decrease by at least 13% in terms of the Pass@1 metric. Furthermore, LCMs perform variously across different programming languages, e.g., the performance gap between Python and C++ reaches as high as 20.9%. ...