The field of software engineering and automation is undergoing a significant transformation with the integration of large language models (LLMs). Recent research has highlighted the potential of LLMs to improve the efficiency and accuracy of various tasks, including business process automation, hardware design automation, software requirements formalisation, software testing and validation, and software development.
A common theme among these areas is the use of LLMs to automate tasks, improve productivity, and enhance quality. In business process automation, LLMs are being used to develop novel frameworks and methodologies for assessing the desirability of process deviations and optimizing the use of process automation. Notable contributions include the development of procedural frameworks for assessing deviation desirability and effective engagement models between technology and process participants.
In hardware design automation, LLMs are being used to generate hardware code, automate reference model design and verification, and enhance the portability of existing code across different instruction set architectures. Innovative approaches, such as Spec2RTL-Agent and Guaranteed Guess, have introduced novel multi-agent collaboration frameworks and ISA-centric transpilation pipelines, respectively.
The field of software requirements formalisation is also benefiting from the adoption of LLMs, with researchers exploring the use of LLMs to automate survey generation, improve traceability of software requirements, and develop new tools and methodologies for formal software verification. Noteworthy papers in this area include A Short Survey on Formalising Software Requirements using Large Language Models and Impact of a Deployed LLM Survey Creation Tool through the IS Success Model.
In software testing and validation, LLMs are being used to automate various testing tasks, such as test generation, oracle generation, and code refactoring. Innovative approaches, such as the use of functional programming and type systems to translate code into formal representations, are being proposed to address the limitations of existing testing methods.
The field of software development is shifting towards collaborative interactions between developers and AI assistants, with a focus on enhancing productivity and code quality. Multi-agent LLM-driven systems and LLMs are being leveraged to automate tasks such as code completion, test case generation, and documentation production. Noteworthy papers in this area include Human-In-The-Loop Software Development Agents, MultiMind, Mind the Metrics, and LLM-as-a-Judge.
Finally, the field of software development and maintenance is witnessing a significant shift towards AI-driven approaches, with the integration of LLMs and human expertise becoming increasingly important in tackling challenges such as reliability, security, and quality. Notable trends include the development of tools and frameworks that analyze commit messages, extract developer rationale, and predict bugs.
Overall, the integration of LLMs in software engineering and automation is revolutionizing various aspects of the field, from business process automation to software development and maintenance. As research in this area continues to evolve, we can expect to see significant improvements in productivity, quality, and efficiency, leading to better software systems and applications.