Using Large Language Models for Legal Decision-Making in Austrian Value-Added Tax Law: An Experimental Investigation of Retrieval-Augmented Generation and Fine-Tuning

This project contains the source code and evaluation setup used for comparing fine-tuning and RAG in the context of Austrian VAT law.

It includes evaluation data, final results, prompt flow definitions for Azure AI Foundry, and metrics related to the fine-tuning process. The training data used for fine-tuning is not included and will not be published.