REPOEXEC: Evaluate Code Generation with a Repository-Level Executable Benchmark
2406.11927
0
0
Abstract
The ability of CodeLLMs to generate executable and functionally correct code at the repository-level scale remains largely unexplored. We introduce RepoExec, a novel benchmark for evaluating code generation at the repository-level scale. RepoExec focuses on three main aspects: executability, functional correctness through automated test case generation with high coverage rate, and carefully crafted cross-file contexts to accurately generate code. Our work explores a controlled scenario where developers specify necessary code dependencies, challenging the model to integrate these accurately. Experiments show that while pretrained LLMs outperform instruction-tuned models in correctness, the latter excel in utilizing provided dependencies and demonstrating debugging capabilities. We also introduce a new instruction-tuned dataset that focuses on code dependencies and demonstrate that CodeLLMs fine-tuned on our dataset have a better capability to leverage these dependencies effectively. RepoExec aims to provide a comprehensive evaluation of code functionality and alignment with developer intent, paving the way for more reliable and applicable CodeLLMs in real-world scenarios. The dataset and source code can be found at~url{https://github.com/FSoft-AI4Code/RepoExec}.
Create account to get full access
Overview
- This paper introduces a new benchmark called RepoExec for evaluating code generation models at the repository level rather than just the class or function level.
- RepoExec aims to better mimic real-world software development by assessing a model's ability to generate complete, executable code repositories that can be integrated into a codebase.
- The paper compares the performance of several state-of-the-art code generation models on the RepoExec benchmark and provides insights into their strengths and limitations.
Plain English Explanation
The paper introduces a new way to evaluate code generation models, which are AI systems that can produce computer programs from natural language descriptions. Existing benchmarks for these models usually focus on generating individual functions or classes of code, but the authors argue this doesn't capture the full complexity of real-world software development.
To address this, they created the RepoExec benchmark, which assesses a model's ability to generate complete, executable code repositories that can be integrated into a larger codebase. This is more representative of the challenges developers face when incorporating AI-generated code into their projects.
The paper then compares the performance of several state-of-the-art code generation models on the RepoExec benchmark. By analyzing the results, the authors gain insights into the strengths and limitations of these models when it comes to producing usable, production-ready code. This information can help guide future research and development efforts in this rapidly advancing field.
Technical Explanation
The paper introduces the RepoExec benchmark, which aims to evaluate code generation models at the repository level rather than just the class or function level. This is an important distinction, as existing benchmarks like ClassLevelCodeGen and DevEval don't fully capture the complexities of integrating AI-generated code into a real-world software project.
The RepoExec benchmark consists of a set of natural language prompts that describe the functionality of a complete code repository. The models are tasked with generating the full repository, including all necessary files, classes, and dependencies, that can be executed and integrated into a larger codebase. This provides a more comprehensive assessment of a model's abilities compared to generating individual code elements.
The paper evaluates the performance of several state-of-the-art code generation models, including Codex and InstructGPT, on the RepoExec benchmark. The results reveal insights into the strengths and limitations of these models, such as their ability to generate complete, executable code repositories versus just individual functions or classes.
Critical Analysis
The RepoExec benchmark represents an important step forward in evaluating code generation models, as it more closely reflects the real-world challenges faced by software developers. However, the paper acknowledges that the benchmark is still limited in scope and may not capture all the nuances of integrating AI-generated code into a complex, production-ready codebase.
Additionally, the paper focuses primarily on evaluating the current state-of-the-art models, but does not delve into the underlying architectural differences or training approaches that may contribute to their performance on the RepoExec benchmark. A more in-depth analysis of these factors could provide valuable insights for future model development.
Furthermore, the paper does not address potential ethical concerns or societal implications of deploying code generation models in real-world software development. As these technologies continue to advance, it will be important to consider issues such as algorithmic bias, security vulnerabilities, and the impact on software engineering job markets.
Conclusion
The RepoExec benchmark introduced in this paper represents a significant advancement in the evaluation of code generation models. By assessing a model's ability to generate complete, executable code repositories, RepoExec provides a more realistic and comprehensive assessment of their capabilities compared to existing benchmarks.
The insights gained from applying RepoExec to several state-of-the-art models can help guide future research and development efforts in this rapidly evolving field. As code generation models continue to improve, the ability to reliably integrate their output into production codebases will be a crucial factor in their real-world adoption and impact.
This summary was produced with help from an AI and may contain inaccuracies - check out the links to read the original source documents!
Related Papers
🛸
Class-Level Code Generation from Natural Language Using Iterative, Tool-Enhanced Reasoning over Repository
Ajinkya Deshpande, Anmol Agarwal, Shashank Shet, Arun Iyer, Aditya Kanade, Ramakrishna Bairi, Suresh Parthasarathy
0
0
LLMs have demonstrated significant potential in code generation tasks, achieving promising results at the function or statement level across various benchmarks. However, the complexities associated with creating code artifacts like classes, particularly within the context of real-world software repositories, remain underexplored. Prior research treats class-level generation as an isolated task, neglecting the intricate dependencies & interactions that characterize real-world software environments. To address this gap, we introduce RepoClassBench, a comprehensive benchmark designed to rigorously evaluate LLMs in generating complex, class-level code within real-world repositories. RepoClassBench includes Natural Language to Class generation tasks across Java, Python & C# from a selection of repositories. We ensure that each class in our dataset not only has cross-file dependencies within the repository but also includes corresponding test cases to verify its functionality. We find that current models struggle with the realistic challenges posed by our benchmark, primarily due to their limited exposure to relevant repository contexts. To address this shortcoming, we introduce Retrieve-Repotools-Reflect (RRR), a novel approach that equips LLMs with static analysis tools to iteratively navigate & reason about repository-level context in an agent-based framework. Our experiments demonstrate that RRR significantly outperforms existing baselines on RepoClassBench, showcasing its effectiveness across programming languages & under various settings. Our findings emphasize the critical need for code-generation benchmarks to incorporate repo-level dependencies to more accurately reflect the complexities of software development. Our work shows the benefits of leveraging specialized tools to enhance LLMs' understanding of repository context. We plan to make our dataset & evaluation harness public.
6/6/2024
DevEval: A Manually-Annotated Code Generation Benchmark Aligned with Real-World Code Repositories
Jia Li, Ge Li, Yunfei Zhao, Yongmin Li, Huanyu Liu, Hao Zhu, Lecheng Wang, Kaibo Liu, Zheng Fang, Lanshen Wang, Jiazheng Ding, Xuanming Zhang, Yuqi Zhu, Yihong Dong, Zhi Jin, Binhua Li, Fei Huang, Yongbin Li
0
0
How to evaluate the coding abilities of Large Language Models (LLMs) remains an open question. We find that existing benchmarks are poorly aligned with real-world code repositories and are insufficient to evaluate the coding abilities of LLMs. To address the knowledge gap, we propose a new benchmark named DevEval, which has three advances. (1) DevEval aligns with real-world repositories in multiple dimensions, e.g., code distributions and dependency distributions. (2) DevEval is annotated by 13 developers and contains comprehensive annotations (e.g., requirements, original repositories, reference code, and reference dependencies). (3) DevEval comprises 1,874 testing samples from 117 repositories, covering 10 popular domains (e.g., Internet, Database). Based on DevEval, we propose repository-level code generation and evaluate 8 popular LLMs on DevEval (e.g., gpt-4, gpt-3.5, StarCoder 2, DeepSeek Coder, CodeLLaMa). Our experiments reveal these LLMs' coding abilities in real-world code repositories. For example, in our experiments, the highest Pass@1 of gpt-4-turbo is only 53.04%. We also analyze LLMs' failed cases and summarize their shortcomings. We hope DevEval can facilitate the development of LLMs in real code repositories. DevEval, prompts, and LLMs' predictions have been released.
5/31/2024
🛸
CodeBenchGen: Creating Scalable Execution-based Code Generation Benchmarks
Yiqing Xie, Alex Xie, Divyanshu Sheth, Pengfei Liu, Daniel Fried, Carolyn Rose
0
0
To facilitate evaluation of code generation systems across diverse scenarios, we present CodeBenchGen, a framework to create scalable execution-based benchmarks that only requires light guidance from humans. Specifically, we leverage a large language model (LLM) to convert an arbitrary piece of code into an evaluation example, 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 revised from code in 367 GitHub repositories taken from the CodeSearchNet dataset. To demonstrate the complexity and 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 the code at https://github.com/Veronicium/CodeBenchGen.
5/9/2024
🎲
Benchmarks and Metrics for Evaluations of Code Generation: A Critical Review
Debalina Ghosh Paul, Hong Zhu, Ian Bayley
0
0
With the rapid development of Large Language Models (LLMs), a large number of machine learning models have been developed to assist programming tasks including the generation of program code from natural language input. However, how to evaluate such LLMs for this task is still an open problem despite of the great amount of research efforts that have been made and reported to evaluate and compare them. This paper provides a critical review of the existing work on the testing and evaluation of these tools with a focus on two key aspects: the benchmarks and the metrics used in the evaluations. Based on the review, further research directions are discussed.
6/19/2024