Execution-based Evaluation for Data Science Code Generation Models
Abstract
Code generation models can benefit data scientists' productivity by automatically generating code from context and text descriptions. An important measure of the modeling progress is whether a model can generate code that can correctly execute to solve the task. However, due to the lack of an evaluation dataset that directly supports execution-based model evaluation, existing work relies on code surface form similarity metrics (e.g., BLEU, CodeBLEU) for model selection, which can be inaccurate. To remedy this, we introduce ExeDS, an evaluation dataset for execution evaluation for data science code generation tasks. ExeDS contains a set of 534 problems from Jupyter Notebooks, each consisting of code context, task description, reference program, and the desired execution output. With ExeDS, we evaluate the execution performance of five state-of-the-art code generation models that have achieved high surface-form evaluation scores. Our experiments show that models with high surface-form scores do not necessarily perform well on execution metrics, and execution-based metrics can better capture model code generation errors. Source code and data can be found at https://github.com/Jun-jie-Huang/ExeDS
Cite
@article{arxiv.2211.09374,
title = {Execution-based Evaluation for Data Science Code Generation Models},
author = {Junjie Huang and Chenglong Wang and Jipeng Zhang and Cong Yan and Haotian Cui and Jeevana Priya Inala and Colin Clement and Nan Duan and Jianfeng Gao},
journal= {arXiv preprint arXiv:2211.09374},
year = {2022}
}
Comments
Accepted by the 4th Workshop on Data Science with Human-in-the-loop (DaSH) at EMNLP 2022