by Bianca Steffes and Nils Wiedemann
Abstract:
Using the task of generating guiding principles for judgments of the German Federal Court of Justice we investigate whether current state of the art large language models can solve a complex legal summarisation task. Our results indicate that prompt engineering is not yet sufficient to solve the task, but fine-tuning already shows promising results. In addition, our results show that models with an increased context window size do not necessarily take the entire input into account equally.
Reference:
Bianca Steffes and Nils Wiedemann: Generating Guiding Principles: Evaluating Large Language Models for Complex German Legal Summaries, In New Frontiers in Artificial Intelligence (Yukiko Nakano, Toyotaro Suzumura, eds.), Springer Nature Singapore, pp. 49–65, 2025.
Bibtex Entry:
@InProceedings{ steffeswiedemannjurisin25,
author = {Steffes, Bianca and Wiedemann, Nils},
editor = {Nakano, Yukiko and Suzumura, Toyotaro},
title = {Generating Guiding Principles: Evaluating Large Language
Models for Complex German Legal Summaries},
booktitle = {New Frontiers in Artificial Intelligence},
year = {2025},
publisher = {Springer Nature Singapore},
address = {Singapore},
pages = {49--65},
abstract = {Using the task of generating guiding principles for
judgments of the German Federal Court of Justice we
investigate whether current state of the art large language
models can solve a complex legal summarisation task. Our
results indicate that prompt engineering is not yet
sufficient to solve the task, but fine-tuning already shows
promising results. In addition, our results show that
models with an increased context window size do not
necessarily take the entire input into account equally.},
isbn = {978-981-96-7071-0}
}