Logo des Repositoriums
 

Combining Retrieval-Augmented Generation and Few-Shot Learning for Model Synthesis of Uncommon DSLs

Kurzanzeige
dc.contributor.authorBaumann, Nils
dc.contributor.authorDiaz, Juan Sebastian
dc.contributor.authorMichael, Judith
dc.contributor.authorNetz, Lukas
dc.contributor.authorNqiri, Haron
dc.contributor.authorReimer, Jan
dc.contributor.authorRumpe, Bernhard
dc.contributor.editorGiese, Holger
dc.contributor.editorRosenthal
dc.contributor.editorKristina
dc.date.accessioned2024-03-12T05:30:27Z
dc.date.available2024-03-12T05:30:27Z
dc.date.issued2024
dc.description.abstractWe introduce a method that empowers large language models (LLMs) to generate models for domain-specific languages (DSLs) for which the LLM has little to no training data on. Common LLMs such as GPT-4, Llama 2, or Bard are trained on publicly available data and thus have the capability to produce models for well-known modeling languages such as PlantUML, however, they perform worse on lesser-known or unpublished DSLs. Previous work focused on the usage of few-shot learning (FSL) to synthesize models but did not address or evaluate the potential of retrieval-augmented generation (RAG) to provide fitting examples for the FSL-based modeling approach. In this work, we propose a toolchain and test each building block individually: We use the MontiCore Sequence Diagram Language, which GPT-4 has minimal training data on, to assess the extent to which FSL enhances the likelihood of synthesizing an accurate model. Additionally, we evaluate how effectively RAG can identify suitable models for user requests and determine whether GPT-4 can distinguish between requests for a specific model and those for general information. We show that RAG and FSL can be used to enable simple model synthesis for uncommon DSLs, as long as there is a fitting knowledge base that can be accessed to provide the needed examples for the FSL approach.en
dc.identifier.doi10.18420/modellierung2024-ws-007
dc.identifier.urihttps://dl.gi.de/handle/20.500.12116/43781
dc.language.isoen
dc.pubPlaceBonn
dc.publisherGesellschaft für Informatik e.V.
dc.relation.ispartofModellierung 2024 Satellite Events
dc.subjectLLMs
dc.subjectRAG
dc.subjectDSLs
dc.subjectFew-Shot Learning
dc.subjectMDSE
dc.titleCombining Retrieval-Augmented Generation and Few-Shot Learning for Model Synthesis of Uncommon DSLsen
dc.typeText/Workshop Paper
gi.conference.date12. - 15. März
gi.conference.locationPotsdam
gi.conference.sessiontitleLLM4Modeling

Dateien

Originalbündel
1 - 1 von 1
Vorschaubild nicht verfügbar
Name:
07_Combining Retrieval-Augmented Generation and Few-Shot Learning for Model Synthesis.pdf
Größe:
561.02 KB
Format:
Adobe Portable Document Format
Herunterladen

Sammlungen

Modellierung 2024 - Workshopband