Can OpenSource beat ChatGPT? -- A Comparative Study of Large Language Models for Text-to-Code Generation

TL;DR

This study compares the performance of five large language models in text-to-code generation tasks, finding that ChatGPT outperforms specialized models like Code Llama in solving programming problems from LeetCode.

Contribution

It provides an empirical evaluation of multiple LLMs for code generation, highlighting ChatGPT's superior performance and analyzing factors affecting output quality.

Findings

ChatGPT outperforms other models in solving LeetCode problems.

Performance decreases with longer prompts and more context.

Error analysis reveals common issues like indentation and incorrect code structure.

Abstract

In recent years, large language models (LLMs) have emerged as powerful tools with potential applications in various fields, including software engineering. Within the scope of this research, we evaluate five different state-of-the-art LLMs - Bard, BingChat, ChatGPT, Llama2, and Code Llama - concerning their capabilities for text-to-code generation. In an empirical study, we feed prompts with textual descriptions of coding problems sourced from the programming website LeetCode to the models with the task of creating solutions in Python. Subsequently, the quality of the generated outputs is assessed using the testing functionalities of LeetCode. The results indicate large differences in performance between the investigated models. ChatGPT can handle these typical programming challenges by far the most effectively, surpassing even code-specialized models like Code Llama. To gain further insights, we measure the runtime as well as the memory usage of the generated outputs and compared them to the other code submissions on Leetcode. A detailed error analysis, encompassing a comparison of the differences concerning correct indentation and form of the generated code as well as an assignment of the incorrectly solved tasks to certain error categories allows us to obtain a more nuanced picture of the results and potential for improvement. The results also show a clear pattern of increasingly incorrect produced code when the models are facing a lot of context in the form of longer prompts.