Short biography
Christoph Mandl is a PhD student at the Chair of Data Science at the University of Passau. He holds a Bachelor’s in Mathematics and a Bachelor’s in Computer Science (University of Passau), as well as a Master’s in Computer Science (University of Passau).
His research focuses on understanding narratives and how they change over time.
Contact
c.mandl <ät> media-bias-research org
References
2026
Kučera, Filip; Mandl, Christoph; Echizen, Isao; Timofte, Radu; Spinde, Timo
SciDef: Automating Definition Extraction from Academic Literature with Large Language Models Miscellaneous
2026.
Abstract | Links | BibTeX | Tags:
@misc{kucera2026scidefautomatingdefinitionextraction,
title = {SciDef: Automating Definition Extraction from Academic Literature with Large Language Models},
author = {Filip Kučera and Christoph Mandl and Isao Echizen and Radu Timofte and Timo Spinde},
url = {https://arxiv.org/abs/2602.05413},
year = {2026},
date = {2026-01-01},
urldate = {2026-01-01},
abstract = {Definitions are the foundation for any scientific work, but with a significant increase in publication numbers, gathering definitions relevant to any keyword has become challenging. We therefore introduce SciDef, an LLM-based pipeline for automated definition extraction. We test SciDef on DefExtra & DefSim, novel datasets of human-extracted definitions and definition-pairs' similarity, respectively. Evaluating 16 language models across prompting strategies, we demonstrate that multi-step and DSPy-optimized prompting improve extraction performance. To evaluate extraction, we test various metrics and show that an NLI-based method yields the most reliable results. We show that LLMs are largely able to extract definitions from scientific literature (86.4% of definitions from our test-set); yet future work should focus not just on finding definitions, but on identifying relevant ones, as models tend to over-generate them.
Code & datasets are available at https://github.com/Media-Bias-Group/SciDef.},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Definitions are the foundation for any scientific work, but with a significant increase in publication numbers, gathering definitions relevant to any keyword has become challenging. We therefore introduce SciDef, an LLM-based pipeline for automated definition extraction. We test SciDef on DefExtra & DefSim, novel datasets of human-extracted definitions and definition-pairs' similarity, respectively. Evaluating 16 language models across prompting strategies, we demonstrate that multi-step and DSPy-optimized prompting improve extraction performance. To evaluate extraction, we test various metrics and show that an NLI-based method yields the most reliable results. We show that LLMs are largely able to extract definitions from scientific literature (86.4% of definitions from our test-set); yet future work should focus not just on finding definitions, but on identifying relevant ones, as models tend to over-generate them.
Code & datasets are available at https://github.com/Media-Bias-Group/SciDef.
Code & datasets are available at https://github.com/Media-Bias-Group/SciDef.
2025
Horych, Tomas; Mandl, Christoph; Ruas, Terry; Greiner-Petter, Andre; Gipp, Bela; Aizawa, Akiko; Spinde, Timo
The Promises and Pitfalls of LLM Annotations in Dataset Labeling: a Case Study on Media Bias Detection Proceedings Article
In: Findings of the 2025 Conference of the The Nations of the Americas Chapter of the Association for Computational Linguistics: NAACL 2025, Association for Computational Linguistics, Albuquerque, USA, 2025.
Abstract | Links | BibTeX | Tags: dataset, lexical bias, LLMs, synthetic annotations
@inproceedings{Horych2025,
title = {The Promises and Pitfalls of LLM Annotations in Dataset Labeling: a Case Study on Media Bias Detection},
author = {Tomas Horych and Christoph Mandl and Terry Ruas and Andre Greiner-Petter and Bela Gipp and Akiko Aizawa and Timo Spinde},
url = {https://media-bias-research.org/wp-content/uploads/2025/01/Horych2025.pdf},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {Findings of the 2025 Conference of the The Nations of the Americas Chapter of the Association for Computational Linguistics: NAACL 2025},
publisher = {Association for Computational Linguistics},
address = {Albuquerque, USA},
abstract = {High annotation costs from hiring or crowd-sourcing complicate the creation of large, high-quality datasets needed for training reliable text classifiers. Recent research suggests using Large Language Models (LLMs) to automate the annotation process, reducing these costs while maintaining data quality. LLMs have shown promising results in annotating downstream tasks like hate speech detection and political framing. Building on the success in these areas, this study investigates whether LLMs are viable for annotating a complex task of media bias detection and whether a downstream media bias classifier can be trained on such data. We create Anno-lexical , the first large-scale dataset for media bias classification with over 48k synthetically annotated examples.
Our classifier fine-tuned on it surpasses all of the annotator LLMs by 5-9% in Mathew’s Cor- relation Coefficient (MCC) and performs close to or outperforms the model trained on human-labeled data when evaluated on two media bias benchmark datasets (BABE and BASIL). This study demonstrates how our approach significantly reduces the cost of dataset creation in the media bias domain and, by extension - the development of the classifiers, while our subse-quent behavioral stress-testing reveals some of its current limitations and trade-offs.},
keywords = {dataset, lexical bias, LLMs, synthetic annotations},
pubstate = {published},
tppubtype = {inproceedings}
}
High annotation costs from hiring or crowd-sourcing complicate the creation of large, high-quality datasets needed for training reliable text classifiers. Recent research suggests using Large Language Models (LLMs) to automate the annotation process, reducing these costs while maintaining data quality. LLMs have shown promising results in annotating downstream tasks like hate speech detection and political framing. Building on the success in these areas, this study investigates whether LLMs are viable for annotating a complex task of media bias detection and whether a downstream media bias classifier can be trained on such data. We create Anno-lexical , the first large-scale dataset for media bias classification with over 48k synthetically annotated examples.
Our classifier fine-tuned on it surpasses all of the annotator LLMs by 5-9% in Mathew’s Cor- relation Coefficient (MCC) and performs close to or outperforms the model trained on human-labeled data when evaluated on two media bias benchmark datasets (BABE and BASIL). This study demonstrates how our approach significantly reduces the cost of dataset creation in the media bias domain and, by extension - the development of the classifiers, while our subse-quent behavioral stress-testing reveals some of its current limitations and trade-offs.
Our classifier fine-tuned on it surpasses all of the annotator LLMs by 5-9% in Mathew’s Cor- relation Coefficient (MCC) and performs close to or outperforms the model trained on human-labeled data when evaluated on two media bias benchmark datasets (BABE and BASIL). This study demonstrates how our approach significantly reduces the cost of dataset creation in the media bias domain and, by extension - the development of the classifiers, while our subse-quent behavioral stress-testing reveals some of its current limitations and trade-offs.