The paper ‘Tasks People Prompt: A Taxonomy of LLM Downstream Tasks in Software Verification and Falsification Approaches’ presents an in-depth analysis of how the software testing and verification communities are utilizing Large Language Models (LLMs) by abstractly architecting their LLM-enabled solutions.

• Studies 80 papers to understand LLM application in software testing and verification.
• Develops a downstream task taxonomy to identify patterns in testing, fuzzing, debugging, vulnerability detection, and more.
• Explores prompt-based solutions and validates the concept of downstream tasks.
• Provides insights into the nature and number of tasks within these solutions.

This research is critical as it demystifies the usage of advanced language models in complex engineering tasks, paving the way for automation and efficiency in software testing. It’s a significant contribution that can guide further research and practical applications in debug automation, program analysis, and other facets of software engineering. Read more