Bugs in Large Language Models Generated Code: An Empirical Study

TL;DR

This empirical study analyzes 333 bugs in code generated by large language models, identifying 10 common bug patterns and validating their significance through a survey, to improve quality assurance in AI-assisted coding.

Contribution

The paper introduces a taxonomy of 10 bug patterns in LLM-generated code and validates their relevance with practitioners, advancing understanding of AI-generated software bugs.

Findings

Identified 10 distinctive bug patterns in LLM-generated code.

Validated bug patterns through a survey with 34 practitioners.

Provides insights for developing quality assurance techniques.

Abstract

Large Language Models (LLMs) for code have gained significant attention recently. They can generate code in different programming languages based on provided prompts, fulfilling a long-lasting dream in Software Engineering (SE), i.e., automatic code generation. Similar to human-written code, LLM-generated code is prone to bugs, and these bugs have not yet been thoroughly examined by the community. Given the increasing adoption of LLM-based code generation tools (e.g., GitHub Copilot) in SE activities, it is critical to understand the characteristics of bugs contained in code generated by LLMs. This paper examines a sample of 333 bugs collected from code generated using three leading LLMs (i.e., CodeGen, PanGu-Coder, and Codex) and identifies the following 10 distinctive bug patterns: Misinterpretations, Syntax Error, Silly Mistake, Prompt-biased code, Missing Corner Case, Wrong Input Type, Hallucinated Object, Wrong Attribute, Incomplete Generation, and Non-Prompted Consideration. The bug patterns are presented in the form of a taxonomy. The identified bug patterns are validated using an online survey with 34 LLM practitioners and researchers. The surveyed participants generally asserted the significance and prevalence of the bug patterns. Researchers and practitioners can leverage these findings to develop effective quality assurance techniques for LLM-generated code. This study sheds light on the distinctive characteristics of LLM-generated code.