Konferenzbeitrag
Chances and Challenges of LLM-based Software Reengineering
Vorschaubild nicht verfügbar
Volltext URI
Dokumententyp
Text/Conference Paper
Zusatzinformation
Datum
2024
Autor:innen
Zeitschriftentitel
ISSN der Zeitschrift
Bandtitel
Verlag
Gesellschaft für Informatik e.V.
Zusammenfassung
Large Language Models (LLMs) have opened up unforeseen new possibilities. They deliver amazing results for complex text-based tasks for which no satisfactory automated solution was available before. This is even more astonishing as they just calculate the most probable subsequent token, given a sequence of tokens. This is also true for software development support: There are various software engineering task for which LLMs show potential. In this paper, we discuss the potential and current shortcomings of LLM-based approaches for selected Software Reengineering tasks with a focus on language translation. All experiments reported in the following were performed with GPT-4.