Software Engineering: How LLMs and Agents Are Changing the Game

Large language models (LLMs) have made a splash in software engineering, especially in code generation and finding coding issues. But they're not perfect. Enter LLM-based agents, which could lead to Artificial General Intelligence (AGI) by making LLMs more autonomous and capable of self-improvement. Experts are still figuring out how to set standards and benchmarks for these agents.

This survey digs into how LLMs and agents are used in software engineering. It looks at six major areas: requirement engineering, code generation, autonomous decision-making, software design, test generation, and maintenance. The study compares the work of LLMs and agents in these fields, checking out how they stack up in tasks, tests, and measurement tools. The review also dives into the models and tests used, giving a thorough look at how effective they are in software engineering. This work aims to inspire future research and push the limits of what LLM-based agents can do in this field.

Actions