CodeBenchGen: Creating Scalable Execution-based Code Generation Benchmarks

TL;DR

CodeBenchGen is a framework that creates scalable, execution-based code generation benchmarks from real-world code, enabling more comprehensive evaluation of code generation systems across diverse scenarios.

Contribution

It introduces a novel method to generate execution-based benchmarks from natural code sources using large language models, expanding evaluation capabilities.

Findings

Created the Exec-CSN dataset with 1,931 examples from GitHub repositories.

81.3% of examples are solvable by humans, indicating practical relevance.

Conducted code generation experiments demonstrating the framework's utility.

Abstract

To adequately test modern code generation systems, evaluation benchmarks must execute and test the code generated by the system. However, these execution and testing requirements have largely limited benchmarks to settings where code is easily executable or has human-written tests. To facilitate evaluation of code generation systems across diverse scenarios, we present CodeBenchGen, a framework to create scalable execution-based benchmarks from naturally occurring code sources. Specifically, we leverage a large language model (LLM) to sandbox arbitrary pieces of code into evaluation examples, including test cases for execution-based evaluation. We illustrate the usefulness of our framework by creating a dataset, Exec-CSN, which includes 1,931 examples involving 293 libraries converted from code in 367 GitHub repositories taken from the Code- SearchNet dataset. To demonstrate the solvability of examples in Exec-CSN, we present a human study demonstrating that 81.3% of the examples can be solved by humans and 61% are rated as "requires effort to solve". We conduct code generation experiments on open-source and proprietary models and analyze the performance of both humans and models. We provide code and data at: https://github.com/yiqingxyq/CodeBenchGen.