Скачать или смотреть Podcast about the paper reviews the application of LLM-based Agents in Software Engineering. part1

From LLMs to LLM-based Agents for Software Engineering: A Survey of Current, Challenges and Future.
This paper reviews the application of large language models (LLMs) and LLM-based agents in software engineering. The authors categorize software engineering into six key themes: requirement engineering and documentation, code generation and software development, autonomous learning and decision making, software design and evaluation, software test generation, and software security and maintenance. For each theme, the paper examines tasks, benchmarks, and evaluation metrics, highlighting the distinctions between LLMs and LLM-based agents. The analysis demonstrates that LLM-based agents have significantly impacted software engineering research and applications, showcasing different emphases compared to traditional LLMs. The paper then provides a comprehensive analysis of the models used in experiments from 117 collected papers and offers insights into dataset and evaluation metric differences. The authors conclude by suggesting that LLM-based agents are a transformative technology with the potential to revolutionize software engineering practices.