MultiPL-E: A Scalable and Polyglot Approach to Benchmarking Neural Code Generation
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 (Chen et al., 2021) and MBPP benchmark (Austin et al., 2021) 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-of-the-art code generation models: Codex (Chen et al., 2021), CodeGen (Nijkamp et al., 2022) and InCoder (Fried et al., 2022). 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.
Tue 5 DecDisplayed time zone: Pacific Time (US & Canada) change
16:00 - 18:00 | Code Search and Text to CodeResearch Papers / Industry Papers / Journal First / Demonstrations at Golden Gate A Chair(s): Miryung Kim University of California at Los Angeles, USA | ||
16:00 15mTalk | [Remote] Self-Supervised Query Reformulation for Code Search Research Papers Yuetian Mao Shanghai Jiao Tong University, Chengcheng Wan East China Normal University, Yuze Jiang Shanghai Jiao Tong University, Xiaodong Gu Shanghai Jiao Tong University Media Attached | ||
16:15 15mTalk | [Remote] Natural Language to Code: How Far are We? Research Papers Shangwen Wang National University of Defense Technology, Mingyang Geng National University of Defense Technology, Bo Lin National University of Defense Technology, Zhensu Sun Singapore Management University, Ming Wen Huazhong University of Science and Technology, Yepang Liu Southern University of Science and Technology, Li Li Beihang University, Tegawendé F. Bissyandé University of Luxembourg, Xiaoguang Mao National University of Defense Technology DOI Pre-print Media Attached | ||
16:30 15mTalk | [Remote] xASTNN: Improved Code Representations for Industrial Practice Industry Papers Zhiwei Xu Tsinghua University, Min Zhou Tsinghua University, Xibin Zhao Tsinghua University, Yang Chen Huazhong University of Science and Technology, Xi Cheng VMware, Hongyu Zhang Chongqing University DOI Media Attached | ||
16:45 7mTalk | [Remote] On the Dual Nature of Necessity in Use of Rust Unsafe Code Industry Papers Yuchen Zhang New York University, USA, Ashish Kundu Cisco Research, Georgios Portokalidis Stevens Institute of Technology, Jun Xu The University of Utah DOI Media Attached | ||
16:53 7mTalk | On Using Information Retrieval to Recommend Machine Learning Good Practices for Software Engineers Demonstrations Laura Cabra-Acela Universidad de Los Andes, Anamaria Mojica-Hanke University of Passau, Universidad de Los Andes, Mario Linares-Vásquez Universidad de los Andes, Steffen Herbold University of Passau Media Attached | ||
17:00 15mTalk | MultiPL-E: A Scalable and Polyglot Approach to Benchmarking Neural Code Generation Journal First Federico Cassano Northeastern University, John Gouwar Northeastern University, Daniel Nguyen Hannover High School, Sydney Nguyen Wellesley College, Luna Phipps-Costin Northeastern University, Donald Pinckney Northeastern University, Ming-Ho Yee Northeastern University, Yangtian Zi Northeastern University, Carolyn Jane Anderson Wellesley College, Molly Q Feldman Oberlin College, Arjun Guha Northeastern University and Roblox, Michael Greenberg Stevens Institute of Technology, Abhinav Jangda Microsoft Research Link to publication Media Attached | ||
17:15 15mTalk | NCQ: Code reuse support for Node.js developers Journal First Brittany Reid The University of Adelaide, Marcelo d'Amorim North Carolina State University, Markus Wagner Monash University, Australia, Christoph Treude University of Melbourne Link to publication DOI Pre-print Media Attached | ||
17:30 15mTalk | Efficient Text-to-Code Retrieval with Cascaded Fast and Slow Transformer Models Research Papers Akhilesh Deepak Gotmare Salesforce Research, Junnan Li Salesforce Research, Shafiq Joty Salesforce Research, Steven C.H. Hoi Salesforce Research Asia Media Attached | ||
17:45 15mTalk | PEM: Representing Binary Program Semantics for Similarity Analysis via A Probabilistic Execution Model Research Papers Xiangzhe Xu Purdue University, Zhou Xuan , Shiwei Feng Purdue University, Siyuan Cheng Purdue University, Yapeng Ye Purdue University, Qingkai Shi The Hong Kong University of Science and Technology, Guanhong Tao Purdue University, Le Yu , Zhuo Zhang Purdue University, Xiangyu Zhang Purdue University Media Attached | ||