MCoNaLa: A Benchmark for Code Generation from Multiple Natural Languages

Code assistance refers to the utilization of various tools, techniques, and models to help developers in the process of software development. As coding tasks become increasingly complex, code assistant plays a pivotal role in enhancing developer productivity, reducing errors, and facilitating a more efficient coding workflow. This assistance can manifest in various forms, including code autocompletion, error detection and correction, code generation, documentation support, and context-aware suggestions. Language models have emerged as integral components of code assistance, offering developers the capability to receive intelligent suggestions, generate code snippets, and enhance overall coding proficiency. In this paper, we propose new hybrid models for code generation by leveraging pre-trained language models BERT, RoBERTa, ELECTRA, and LUKE with the Marian Causal Language Model. Selecting these models based on their strong performance in various natural language processing tasks. We evaluate the performance of these models on two datasets CoNaLa and DJANGO and compare them to existing state-of-the-art models. We aim to investigate the potential of pre-trained transformer language models to revolutionize code generation, offering improved precision and efficiency in navigating complex coding scenarios. Additionally, conducting error analysis and refining the generated code. Our results show that these models, when combined with the Marian Decoder, significantly improve code generation accuracy and efficiency. Notably, the RoBERTaMarian model achieved a maximum BLEU score of 35.74 and an exact match accuracy of 13.8% on CoNaLa, while LUKE-Marian attained a BLEU score of 89.34 and an exact match accuracy of 78.50% on DJANGO. Implementation of this work is available at https://github.com/AhmedSSoliman/Leveraging-Pretrained-Language-Models-for-Code-Generation.

show abstract

MultiPL-E: A Scalable and Polyglot Approach to Benchmarking Neural Code Generation

Federico

Gouwar

Nguyen

et al. 2023

IIEEE Trans. Software Eng.

View full text Add to dashboard Cite

Large language models have demonstrated the ability to generate both natural language and programming language text. Although contemporary code generation models are trained on corpora with several programming languages, they are tested using benchmarks that are typically monolingual. The most widely used code generation benchmarks only target Python, so there is little quantitative evidence of how code generation models perform on other programming languages. We propose MultiPL-E, a system for translating unit test-driven code generation benchmarks to new languages. We create the first massively multilingual code generation benchmark by using MultiPL-E to translate two popular Python code generation benchmarks to 18 additional programming languages.We use MultiPL-E to extend the HumanEval benchmark [1] and MBPP benchmark [2] to 18 languages that encompass a range of programming paradigms and popularity. Using these new parallel benchmarks, we evaluate the multi-language performance of three state-ofthe-art code generation models: Codex [1], CodeGen [3] and InCoder [4]. We find that Codex matches or even exceeds its performance on Python for several other languages. The range of programming languages represented in MultiPL-E allow us to explore the impact of language frequency and language features on model performance. Finally, the MultiPL-E approach of compiling code generation benchmarks to new programming languages is both scalable and extensible, making it straightforward to evaluate new models, benchmarks, and languages.4. These source-to-source compilers are sometimes called transpilers.

show abstract

MCoNaLa: A Benchmark for Code Generation from Multiple Natural Languages

Cited by 6 publications

References 40 publications

Deep learning for code generation: a survey

Deep learning for code generation: a survey

Leveraging pre-trained language models for code generation

MultiPL-E: A Scalable and Polyglot Approach to Benchmarking Neural Code Generation

Contact Info

Product

Resources

About