Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Personality-Guided Code Generation Using Large Language Models

Yaoqi Guo, Zhenpeng Chen, Jie M. Zhang, Yang Liu, Yun Ma

TL;DR

This work investigates whether assigning task-specific programmer personalities can improve large language model (LLM)–driven code generation. It introduces a two-stage pipeline in which GPT-4o generates an MBTI personality for each coding task, after which a target LLM codes as that personality, evaluated across seven LLMs and four datasets. Results show that personality guidance enhances code-generation accuracy in a majority of LLM-dataset pairs, with several cases exceeding 5%–10% gains and a notable 12.9% improvement in at least one setting; gains are amplified when combined with Chain-of-Thought prompting. The study highlights the practical value of personality diversity and prompts designs, offering open-source code and data to enable reproducibility and further exploration.

Abstract

Code generation, the automatic creation of source code from natural language descriptions, has garnered significant attention due to its potential to streamline software development. Inspired by research that links task-personality alignment with improved development outcomes, we conduct an empirical study on personality-guided code generation using large language models (LLMs). Specifically, we investigate how emulating personality traits appropriate to the coding tasks affects LLM performance. We extensively evaluate this approach using seven widely adopted LLMs across four representative datasets. Our results show that personality guidance significantly enhances code generation accuracy, with improved pass rates in 23 out of 28 LLM-dataset combinations. Notably, in 11 cases, the improvement exceeds 5%, and in 5 instances, it surpasses 10%, with the highest gain reaching 12.9%. Additionally, personality guidance can be easily integrated with other prompting strategies to further boost performance. We open-source our code and data at https://github.com/IanWalls/Persona-Code.

Personality-Guided Code Generation Using Large Language Models

TL;DR

This work investigates whether assigning task-specific programmer personalities can improve large language model (LLM)–driven code generation. It introduces a two-stage pipeline in which GPT-4o generates an MBTI personality for each coding task, after which a target LLM codes as that personality, evaluated across seven LLMs and four datasets. Results show that personality guidance enhances code-generation accuracy in a majority of LLM-dataset pairs, with several cases exceeding 5%–10% gains and a notable 12.9% improvement in at least one setting; gains are amplified when combined with Chain-of-Thought prompting. The study highlights the practical value of personality diversity and prompts designs, offering open-source code and data to enable reproducibility and further exploration.

Abstract

Code generation, the automatic creation of source code from natural language descriptions, has garnered significant attention due to its potential to streamline software development. Inspired by research that links task-personality alignment with improved development outcomes, we conduct an empirical study on personality-guided code generation using large language models (LLMs). Specifically, we investigate how emulating personality traits appropriate to the coding tasks affects LLM performance. We extensively evaluate this approach using seven widely adopted LLMs across four representative datasets. Our results show that personality guidance significantly enhances code generation accuracy, with improved pass rates in 23 out of 28 LLM-dataset combinations. Notably, in 11 cases, the improvement exceeds 5%, and in 5 instances, it surpasses 10%, with the highest gain reaching 12.9%. Additionally, personality guidance can be easily integrated with other prompting strategies to further boost performance. We open-source our code and data at https://github.com/IanWalls/Persona-Code.

Paper Structure

This paper contains 27 sections, 2 equations, 5 figures, 10 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. LLM for Code Generation
  4. Prompt Engineering for Code Generation
  5. Personality of LLM
  6. Personality-Guided Code Generation
  7. Experimental Setup
  8. Research Questions (RQs)
  9. Datasets
  10. LLMs Used for Code Generation
  11. Evaluation Metric
  12. Results
  13. RQ1: Effectiveness
  14. RQ2: Influencing Factors
  15. Model Perspective Analysis
  16. ...and 12 more sections

Figures (5)

  • Figure 1: Workflow of personality-guided code generation
  • Figure 2: Distribution of MBTI types generated by GPT-4o for each dataset
  • Figure 3: Venn diagram illustrating the tasks solved by INTJ versus ISTJ types
  • Figure 4: Personality example recommended by GPT-4o
  • Figure 5: Few-shot and CoT prompt structures