Institutsseminar/2026-04-24

Aus SDQ-Institutsseminar
Termin (Alle Termine)
Datum Freitag, 24. April 2026
Uhrzeit 14:00 – 15:45 Uhr (Dauer: 105 min)
Ort Raum 010 (Gebäude 50.34)
Prüfer/in Anne Koziolek
Webkonferenz
Vorheriger Termin Fr 17. April 2026
Nächster Termin Fr 22. Mai 2026
Die Dauer dieses Termins beträgt derzeit 105 Minuten. Bitte ggf. einen weiteren Raum reservieren und den Termin auf zwei Räume aufteilen. Dazu unter Termine eine zusätzliche Terminseite anlegen und die Vorträge neu zuweisen. Falls der Termin dennoch stattfinden soll, bitte im Gruppenkalender die Dauer anpassen und überprüfen, ob der Raum für die ganze Zeit gebucht ist.

Termin in Kalender importieren: iCal (Download)

Vorträge

Formalizing Requirements using Large Language Models
Vortragende(r) Lars Neidiger
Vortragstyp Bachelorarbeit
Betreuer(in) Tobias Hey
Vortragssprache Deutsch
Vortragsmodus in Präsenz
Kurzfassung To ensure that a system satisfies its requirements, Model Checking can be used. To be able to perform Model Checking, the requirements first need to be translated from Natural Language (NL) into a formal logic, like Computation Tree Logic (CTL). Automating this translation task would be beneficial, since it is a time consuming and error-prone process. Existing research has shown that Large Language Models ( LLMs) can at least partly automate this task. A notable gap in existing research is the lack of an evaluation of how an LLM performs at translating the requirements of previously unseen projects into CTL. To fill this gap, I first annotated an existing dataset of NL requirement to CTL translations with the source project of each requirement. Subsequently, I used this dataset to evaluate how different LLMs perform at this translation task on unseen projects using different prompting techniques. I also evaluated the performance of one fine-tuned LLM on this task. The best of the approaches translated 73% of NL requirements accurately, which shows that, while not perfect, an LLM can be helpful in aiding with the translation of NL requirements to CTL, even on unseen projects.
Inferring the execution semantics of executable Domain Specific Languages
Vortragende(r) Yannik Schmid
Vortragstyp Masterarbeit
Betreuer(in) Erik Burger
Vortragssprache Englisch
Vortragsmodus in Präsenz
Kurzfassung Developing domain-specific languages (DSLs), particularly executable DSLs (xDSLs), is a complex task. This complexity often reduces the integration of do-

main experts during the development of a DSL, which can result in DSLs and tools that do not fulfil the user’s requirements. To reduce the complexity of defining the execution semantics of xDSLs, this work proposes an approach to infer the operational execution semantics of a DSL in the form of OCL based on exam- ples that could be provided by a non-language expert. The inference is done using genetic programming and is experimentally evaluated. The results indicate that the approach of this work can be used to infer simple invariants but needs further improvements to possibly infer more complex invariants.

Semantic Tokenization in Source Code Plagiarism Detection
Vortragende(r) Simon Wessel
Vortragstyp Bachelorarbeit
Betreuer(in) Robin Maisch
Vortragssprache Deutsch
Vortragsmodus in Präsenz
Kurzfassung TBD

Neuen Vortrag erstellen

Bitte melde Dich mit Deinen KIT-Zugangsdaten an. Falls Du trotz Anmeldung diese Meldung siehst, bitte eine/-n Wissensmanager/-in darum, Dich zur richtigen Gruppe hinzuzufügen.

Hinweise