Adrian Paschke
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Adrian Paschke has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
- Sefika Efeoglu
Freie Universität Berlin, Germany
2 shared publications - Jan Fillies
Freie Universität Berlin
1 shared publication - Kumar Manas
Freie Universität Berlin
1 shared publication - Sonja Schimmler
Fraunhofer FOKUS, Berlin, Germany
1 shared publication
Latest Publications
- Open Access
- 6 January 2026
- Published by MDPI AG in Preprints.org
Large Language Models for Continual Relation Extraction
Real-world data streams, such as news articles and social media posts, are dynamic and nonstationary, creating challenges for real-time structured representation via knowledge graphs, where relation extraction is a key component. Continual relation extraction (CRE) addresses this setting by incrementally learning new relations while preserving previously acquired knowledge. This work investigates the use of pretrained language models for CRE, focusing on large language models (LLMs) and the effectiveness of memory replay in mitigating forgetting. We evaluated decoder-only models and an encoder-decoder model on TACRED and FewRel in English. Our results show that memory replay is most beneficial for smaller instruction-tuned models (e.g., Flan-T5 Base) and base models such as Llama2-7B-hf. In contrast, the remaining instruction-tuned models in this work do not benefit from memory replay, yet some, like Mistral-7B, already achieve higher accuracies without it and surpass prior methods. We further observed that Llama models in this work are more prone to hallucinations. To the best of our knowledge, this work provides the first reproducible benchmarks for LLMs in CRE. It offers a novel analysis of knowledge retention and hallucination behavior—dimensions that have not been systematically studied in earlier research.
- 6 January 2026
- Published by MDPI AG in Preprints.org
Open AccessLarge Language Models for Continual Relation Extraction
Real-world data streams, such as news articles and social media posts, are dynamic and nonstationary, creating challenges for real-time structured representation via knowledge graphs, where relation extraction is a key component. Continual relation extraction (CRE) addresses this setting by incrementally learning new relations while preserving previously acquired knowledge. This work investigates the use of pretrained language models for CRE, focusing on large language models (LLMs) and the effectiveness of memory replay in mitigating forgetting. We evaluated decoder-only models and an encoder-decoder model on TACRED and FewRel in English. Our results show that memory replay is most beneficial for smaller instruction-tuned models (e.g., Flan-T5 Base) and base models such as Llama2-7B-hf. In contrast, the remaining instruction-tuned models in this work do not benefit from memory replay, yet some, like Mistral-7B, already achieve higher accuracies without it and surpass prior methods. We further observed that Llama models in this work are more prone to hallucinations. To the best of our knowledge, this work provides the first reproducible benchmarks for LLMs in CRE. It offers a novel analysis of knowledge retention and hallucination behavior—dimensions that have not been systematically studied in earlier research.
- Open Access
- 12 March 2025
- Published by Springer Nature in AI and Ethics
- Vol. 5 (4) , 3953-3965
Youth language and emerging slurs: tackling bias in BERT-based hate speech detection
With the increasing presence of adolescents and children online, it is crucial to evaluate algorithms designed to protect them from physical and mental harm. This study measures the bias introduced by emerging slurs found in youth language on existing BERT-based hate speech detection models. The research establishes a novel framework to identify language bias within trained networks, introducing a technique to detect emerging hate phrases and evaluate the unintended bias associated with them. As a result, three bias test sets are constructed: one for emerging hate speech terms, another for established hate terms, and one to test for overfitting. Based on these test sets, three scientific and one commercial hate speech detection models are assessed and compared. For comprehensive evaluation, the research introduces a novel Youth Language Bias Score. Finally, the study applies fine-tuning as a mitigation strategy for youth language bias, rigorously testing and evaluating the newly trained classifier. To summarize, the research introduces a novel framework for bias detection, highlights the influence of adolescent language on classifier performance in hate speech classification, and presents the first-ever hate speech classifier specifically trained for online youth language. This study focuses only on slurs in hateful speech, offering a foundational perspective for the field.
- 12 March 2025
- Published by Springer Nature in AI and Ethics
- Vol. 5 (4) , 3953-3965
Open AccessYouth language and emerging slurs: tackling bias in BERT-based hate speech detection
With the increasing presence of adolescents and children online, it is crucial to evaluate algorithms designed to protect them from physical and mental harm. This study measures the bias introduced by emerging slurs found in youth language on existing BERT-based hate speech detection models. The research establishes a novel framework to identify language bias within trained networks, introducing a technique to detect emerging hate phrases and evaluate the unintended bias associated with them. As a result, three bias test sets are constructed: one for emerging hate speech terms, another for established hate terms, and one to test for overfitting. Based on these test sets, three scientific and one commercial hate speech detection models are assessed and compared. For comprehensive evaluation, the research introduces a novel Youth Language Bias Score. Finally, the study applies fine-tuning as a mitigation strategy for youth language bias, rigorously testing and evaluating the newly trained classifier. To summarize, the research introduces a novel framework for bias detection, highlights the influence of adolescent language on classifier performance in hate speech classification, and presents the first-ever hate speech classifier specifically trained for online youth language. This study focuses only on slurs in hateful speech, offering a foundational perspective for the field.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 1-7
Fine-Tuning Large Language Models for Relation Extraction within a Retrieval-Augmented Generation Framework
Sefika Efeoglu, Adrian Paschke. Proceedings of the 1st Joint Workshop on Large Language Models and Structure Modeling (XLLM 2025). 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 1-7
Open AccessFine-Tuning Large Language Models for Relation Extraction within a Retrieval-Augmented Generation Framework
Sefika Efeoglu, Adrian Paschke. Proceedings of the 1st Joint Workshop on Large Language Models and Structure Modeling (XLLM 2025). 2025.
- 15 October 2023
- Published by Springer Nature
- p. 175-189
Semantic Role Assisted Natural Language Rule Formalization for Intelligent Vehicle
- 15 October 2023
- Published by Springer Nature
- p. 175-189
Semantic Role Assisted Natural Language Rule Formalization for Intelligent Vehicle