Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Survey of Vibe Coding with Large Language Models

Yuyao Ge, Lingrui Mei, Zenghao Duan, Tianhao Li, Yujia Zheng, Yiwei Wang, Lexin Wang, Jiayu Yao, Tianyu Liu, Yujun Cai, Baolong Bi, Fangda Guo, Jiafeng Guo, Shenghua Liu, Xueqi Cheng

TL;DR

This survey formalizes Vibe Coding as a triadic, human–project–agent discipline and introduces a Constrained MD P model to capture the dynamic interactions in AI-assisted software development. It compiles a taxonomy of five development models—Unconstrained Automation, Iterative Conversational Collaboration, Planning-Driven, Test-Driven, and Context-Enhanced—and highlights how context engineering and robust development environments drive success beyond raw agent capability. The work synthesizes the ecosystem across LLM foundations, coding agents, development environments, and multi-source feedback, offering practical guidelines and frameworks for implementation. It also delineates key challenges in technical infrastructure, security, and human factors, providing a roadmap for scalable, trustworthy, and human-centered AI-augmented software engineering.

Abstract

The advancement of large language models (LLMs) has catalyzed a paradigm shift from code generation assistance to autonomous coding agents, enabling a novel development methodology termed "Vibe Coding" where developers validate AI-generated implementations through outcome observation rather than line-by-line code comprehension. Despite its transformative potential, the effectiveness of this emergent paradigm remains under-explored, with empirical evidence revealing unexpected productivity losses and fundamental challenges in human-AI collaboration. To address this gap, this survey provides the first comprehensive and systematic review of Vibe Coding with large language models, establishing both theoretical foundations and practical frameworks for this transformative development approach. Drawing from systematic analysis of over 1000 research papers, we survey the entire vibe coding ecosystem, examining critical infrastructure components including LLMs for coding, LLM-based coding agent, development environment of coding agent, and feedback mechanisms. We first introduce Vibe Coding as a formal discipline by formalizing it through a Constrained Markov Decision Process that captures the dynamic triadic relationship among human developers, software projects, and coding agents. Building upon this theoretical foundation, we then synthesize existing practices into five distinct development models: Unconstrained Automation, Iterative Conversational Collaboration, Planning-Driven, Test-Driven, and Context-Enhanced Models, thus providing the first comprehensive taxonomy in this domain. Critically, our analysis reveals that successful Vibe Coding depends not merely on agent capabilities but on systematic context engineering, well-established development environments, and human-agent collaborative development models.

A Survey of Vibe Coding with Large Language Models

TL;DR

This survey formalizes Vibe Coding as a triadic, human–project–agent discipline and introduces a Constrained MD P model to capture the dynamic interactions in AI-assisted software development. It compiles a taxonomy of five development models—Unconstrained Automation, Iterative Conversational Collaboration, Planning-Driven, Test-Driven, and Context-Enhanced—and highlights how context engineering and robust development environments drive success beyond raw agent capability. The work synthesizes the ecosystem across LLM foundations, coding agents, development environments, and multi-source feedback, offering practical guidelines and frameworks for implementation. It also delineates key challenges in technical infrastructure, security, and human factors, providing a roadmap for scalable, trustworthy, and human-centered AI-augmented software engineering.

Abstract

The advancement of large language models (LLMs) has catalyzed a paradigm shift from code generation assistance to autonomous coding agents, enabling a novel development methodology termed "Vibe Coding" where developers validate AI-generated implementations through outcome observation rather than line-by-line code comprehension. Despite its transformative potential, the effectiveness of this emergent paradigm remains under-explored, with empirical evidence revealing unexpected productivity losses and fundamental challenges in human-AI collaboration. To address this gap, this survey provides the first comprehensive and systematic review of Vibe Coding with large language models, establishing both theoretical foundations and practical frameworks for this transformative development approach. Drawing from systematic analysis of over 1000 research papers, we survey the entire vibe coding ecosystem, examining critical infrastructure components including LLMs for coding, LLM-based coding agent, development environment of coding agent, and feedback mechanisms. We first introduce Vibe Coding as a formal discipline by formalizing it through a Constrained Markov Decision Process that captures the dynamic triadic relationship among human developers, software projects, and coding agents. Building upon this theoretical foundation, we then synthesize existing practices into five distinct development models: Unconstrained Automation, Iterative Conversational Collaboration, Planning-Driven, Test-Driven, and Context-Enhanced Models, thus providing the first comprehensive taxonomy in this domain. Critically, our analysis reveals that successful Vibe Coding depends not merely on agent capabilities but on systematic context engineering, well-established development environments, and human-agent collaborative development models.
Paper Structure (122 sections, 6 equations, 8 figures, 5 tables)

This paper contains 122 sections, 6 equations, 8 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Related Surveys
  4. Foundational LLM.
  5. In-Context Learning.
  6. Post-Training.
  7. Agent Systems.
  8. Preliminary
  9. Reinforcement Learning for Code Generation.
  10. Autonomous Coding Agent Systems.
  11. Function Calling.
  12. Supervised Fine-Tuning.
  13. Vibe Coding: The Engineering of Managing Coding Agents
  14. Definition of Vibe Coding
  15. Formalization of the Triadic Relationship.
  16. ...and 107 more sections

Figures (8)

  • Figure 1: The taxonomy of Vibe Coding is categorized into large language model foundations, coding agent architectures, development environments, and feedback mechanisms. Each area encompasses specific techniques and frameworks that collectively advance the systematic integration of LLMs and agents into intelligent and collaborative software development workflows.
  • Figure 2: Vibe Coding Overview. A triadic collaboration among developer, coding agent, and project, where iterative instruction–feedback loops enable context-aware coding and result-oriented review.
  • Figure 3: Vibe Coding Evolution Timeline: A comprehensive visualization of the development trajectory of Vibe Coding implementations from 2022 to 2025, showing the evolution from foundational Coding LLMs to sophisticated coding agent systems, execute environments and Feedback mechanism.
  • Figure 4: Large Language Model for Coding: Overview of the training lifecycle encompassing Data Foundation, Pre-training, and Post-training stages, with core capabilities including Code Generation, Code Understanding, Bug Detection, and Code Optimization.
  • Figure 5: Coding Agent Architecture: Overview of core components comprising Cognitive System, Memory Mechanism, and Tool Integration, with fundamental capabilities including Planning, Action execution, and multi-agent Collaboration for coordinated software development tasks.
  • ...and 3 more figures