@inproceedings{ArzbergerTavakoliKolagari2026, author = {Arzberger, Alexandra and Tavakoli Kolagari, Ramin}, title = {Conceptional Guidelines for Robustness Evaluation of Autonomous Agents}, series = {SE2026 - Companion Proceedings}, booktitle = {SE2026 - Companion Proceedings}, publisher = {Gesellschaft f{\"u}r Informatik}, address = {Bonn}, doi = {10.18420/se2026-ws_01}, pages = {12}, year = {2026}, abstract = {Autonomous vehicles rely on high-quality sensor data, such as from camera and Light Detection and Ranging (LiDAR) for their driving decisions. Object detection and motion planning are core tasks in autonomous driving, both requiring robustness against diverse sensor perturbations. While robustness evaluation for object detection is well studied, far fewer approaches exist for end-to-end (E2E) motion planning agents. This paper outlines key differences between object detection and E2E motion planning and discusses resulting challenges for robustness evaluation. We present an empirical study comparing the robustness of the object detection system LoGoNet and the E2E agent TransFuser++, both using camera and LiDAR inputs, under LiDAR-only perturbations. Based on our findings, we propose requirements and a parameterization scheme for perturbations to guide the design of future robustness evaluation frameworks for autonomous agents.}, language = {en} } @techreport{Stappert2026, author = {Stappert, Friedhelm}, title = {On the Examination Timetabling Problem at TH N{\"u}rnberg, Dept. of Computer Science}, doi = {10.34646/thn/ohmdok-4690}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:92-opus4-46904}, pages = {26}, year = {2026}, abstract = {This report examines the Examination Timetabling Problem (ETP) at the Department of Computer Science (IN) at Technische Hochschule N{\"u}rnberg (THN), focusing on the specific constraints and challenges of the institution. The report surveys various algorithms and approaches applied to solve the ETP. The Step Counting Hill Climbing (SCHC) algorithm, originally proposed by Bykov and Petrovic, is highlighted as a particularly effective solution. The report details the implementation and enhancements of the SCHC algorithm at IN, which has been successfully applied to the department's timetabling needs. Experimental results demonstrate that the SCHC algorithm, with its simplicity and efficiency, produces high-quality timetables within minutes, comparable to manually created schedules. The report also discusses the limitations of integer linear programming for real-world ETPs due to its exponential complexity. The findings suggest that the SCHC algorithm is well-suited for the ETP at IN, with potential for further optimisation to improve speed and solution quality.}, language = {en} } @article{MonzKroenerFrank2025, author = {Monz, Alexander and Kr{\"o}ner, Alexander and Frank, Johannes}, title = {Benchmark und prototypische Umsetzung eines Raumbelegungssystems}, series = {Schriftenreihe der Technischen Hochschule N{\"u}rnberg: Lernen durch Forschen - Forschungsprojekte mit Bachelorstudierenden}, journal = {Schriftenreihe der Technischen Hochschule N{\"u}rnberg: Lernen durch Forschen - Forschungsprojekte mit Bachelorstudierenden}, number = {2023/2024}, address = {N{\"u}rnberg}, issn = {1867-4585}, pages = {216 - 220}, year = {2025}, subject = {Digitalisierung}, language = {de} } @inproceedings{SteigerwaldBienleinBurghardtetal.2025, author = {Steigerwald, Philipp and Bienlein, Nico and Burghardt, Jennifer and Stieler, Mara and Lehmann, Robert and Albrecht, Jens}, title = {CAIA in Practice: Field Evaluation of an AI-Assisted Support System for Text-Based Online Counselling}, series = {2025 IEEE 37th International Conference on Tools with Artificial Intelligence (ICTAI)}, booktitle = {2025 IEEE 37th International Conference on Tools with Artificial Intelligence (ICTAI)}, publisher = {IEEE}, doi = {10.1109/ICTAI66417.2025.00214}, pages = {1476 -- 1483}, year = {2025}, abstract = {Rising global demand for mental health support creates significant service delivery challenges, with asynchronous email counselling serving as a crucial low-threshold channel for accessing care. This paper presents CAIA, a co-designed AIbased tool suite that demonstrates responsible AI integration into counselling practice through seven LLM-driven functions enhanced by retrieval-augmented generation. A field evaluation involved 34 professional counsellors conducting authentic sessions with trained student counsellees (36 threads, 321 messages, 1,257 AI outputs). User behaviour analysis confirms substantial adoption, revealing that professional autonomy and information accuracy are decisive for sustained acceptance, with counsellors particularly valuing interpretive functionalities that provide new perspectives and stimulate professional reflection.}, language = {en} } @incollection{SteigerwaldAlbrecht2025, author = {Steigerwald, Philipp and Albrecht, Jens}, title = {Enhancing Psychosocial Counselling with AI: A Multifaceted Support System for Professionals}, series = {Frontiers of Artificial Intelligence, Ethics and Multidisciplinary Applications}, booktitle = {Frontiers of Artificial Intelligence, Ethics and Multidisciplinary Applications}, publisher = {Springer Nature}, address = {Singapore}, isbn = {9789819679447}, issn = {2731-8125}, doi = {10.1007/978-981-96-7945-4_35}, pages = {597 -- 611}, year = {2025}, abstract = {The expanding field of psychosocial online email counselling has created a demand for advanced tools to support counsellors in managing diverse client interactions. This paper introduces CAIA (Counsellor Artificial Intelligence Assistant), a comprehensive AI-driven system currently under active development, designed to condense and present information to counsellors in a structured manner. CAIA integrates several AI-based features, including Automatic Subject Generation, Case Summarisation, Timeline Management, Social Network Graph creation, and Case Classification. The features are designed to assist the counsellors, allowing them to engage more deeply with each case while ensuring that the client's needs remain paramount. Each feature presented in the paper addresses a specific problem in psychosocial online counselling, proposes a solution, discusses a technical implementation approach, and highlights the challenges encountered. Additionally, the paper discusses ethical considerations associated with an AI integration.}, language = {en} } @inproceedings{SteigerwaldAlbrecht2025, author = {Steigerwald, Philipp and Albrecht, Jens}, title = {Comparing Large Language Models for Automated Subject Line Generation in e-Mental Health: A Performance Study}, series = {Proceedings of the 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health}, volume = {2025}, booktitle = {Proceedings of the 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health}, publisher = {SCITEPRESS - Science and Technology Publications}, doi = {10.5220/0013294100003938}, pages = {70 -- 77}, year = {2025}, abstract = {Large Language Models (LLMs) have the potential to enhance e-mental health and psychosocial e-mail counselling by automating tasks such as generating concise and relevant subject lines for client communications. However, concerns regarding accuracy, reliability, data privacy and resource efficiency persist. This study investigates the performance of several LLMs in generating subject lines for e-mail threads, yielding a total of 253 generated subjects. Each subject line was assessed by six raters, including five counselling professionals and one AI system, using a three-category quality scale (Good, Fair, Poor). The results show that LLMs can generally produce concise subject lines considered helpful by experts. While GPT-4o and GPT-3.5 Turbo outperformed other models, their use is restricted in mental health settings due to data protection concerns, making the evaluation of open-source models crucial. Among open-source models, SauerkrautLM LLama 3 70b (4-bit) and SauerkrautLM Mixtral 8x7b (both 8-bit and 4-bit versions) delivered promising results with potential for further development. In contrast, models with lower parameter counts produced predominantly poor outputs.}, language = {en} } @article{LeeBokLeeWagneretal.2025, author = {Lee, Seanie and Bok Lee, Dong and Wagner, Dominik and Kang, Minki and Seong, Haebin and Bocklet, Tobias and Lee, Juho and Hwang, Sung Ju}, title = {SafeRoute: Adaptive Model Selection for Efficient and Accurate Safety Guardrails in Large Language Models,}, series = {Findings of the Association for Computational Linguistics: ACL 2025}, journal = {Findings of the Association for Computational Linguistics: ACL 2025}, publisher = {Association for Computational Linguistics}, address = {Vienna, Austria}, pages = {2053-2069}, year = {2025}, abstract = {Deploying large language models (LLMs) in real-world applications requires robust safety guard models to detect and block harmful user prompts. While large safety guard models achieve strong performance, their computational cost is substantial. To mitigate this, smaller distilled models are used, but they often underperform on "hard" examples where the larger model provides accurate predictions. We observe that many inputs can be reliablyhandled by the smaller model, while only a small fraction require the larger model's capacity. Motivated by this, we propose SafeRoute, a binary router that distinguishes hard examples from easy ones. Our method selectively applies the larger safety guard model to the data that the router considers hard, improving efficiency while maintaining accuracy compared to solely using the larger safety guard model. Experimental results on multiple benchmark datasets demonstrate that our adaptive model selection significantly enhances the trade-off between computational cost and safety performance, outperforming relevant baselines.}, language = {en} } @inproceedings{BraunWitzlErzigkeitetal.2025, author = {Braun, Franziska and Witzl, Christopher and Erzigkeit, Andreas and Lehfeld, Hartmut and Hillemacher, Thomas and Bocklet, Tobias and Riedhammer, Korbinian}, title = {Pitfalls and Limits in Automatic Dementia Assessment}, series = {Interspeech 2025}, booktitle = {Interspeech 2025}, publisher = {ISCA}, issn = {2958-1796}, doi = {10.21437/Interspeech.2025-2280}, pages = {5663 -- 5667}, year = {2025}, abstract = {Current work on speech-based dementia assessment focuses on either feature extraction to predict assessment scales, or on the automation of existing test procedures. Most research uses public data unquestioningly and rarely performs a detailed error analysis, focusing primarily on numerical performance. We perform an in-depth analysis of an automated standardized dementia assessment, the Syndrom-Kurz-Test. We find that while there is a high overall correlation with human annotators, due to certain artifacts, we observe high correlations for the severely impaired individuals, which is less true for the healthy or mildly impaired ones. Speech production decreases with cognitive decline, leading to overoptimistic correlations when test scoring relies on word naming. Depending on the test design, fallback handling introduces further biases that favor certain groups. These pitfalls remain independent of group distributions in datasets and require differentiated analysis of target groups.}, language = {en} } @inproceedings{FischerTavakoliKolagari2024, author = {Fischer, Alexander and Tavakoli Kolagari, Ramin}, title = {Towards Robust Autonomous Vehicle Software: Extending SAM for a Holistic Security Analysis}, series = {2024 8th International Conference on System Reliability and Safety (ICSRS)}, booktitle = {2024 8th International Conference on System Reliability and Safety (ICSRS)}, publisher = {IEEE}, isbn = {979-8-3503-5450-8}, doi = {10.1109/ICSRS63046.2024.10927601}, pages = {578 -- 586}, year = {2024}, abstract = {The Security Abstraction Model (SAM) is a meta-model that enables comprehensive security modeling and does not only consider technological aspects, but also takes into consideration the influence of social engineering. This plays an important role by exploiting vulnerabilities in human behavior and thus represents a dimension that must not be neglected in possible attacks on vehicle systems. This work is dedicated to the advancement of the Security Abstraction Model (SAM) with the aim of enabling improved modelling and analysis of vulnerabilities in the context of software and system security; the relevant standards and best practices of the international security literature for embedded systems are harmonized in one metamodel. By extending SAM, this research enables a more nuanced understanding and representation of vulnerabilities, attacks and scenarios, improving the effectiveness of security analysis and remediation strategies. The integration of different perspectives, including technological, social and behavioral factors, underlines the comprehensive nature of the proposed model. This interdisciplinary approach recognizes the complexity of security threats in autonomous vehicle systems and enables stakeholders to develop more effective countermeasures. This work presents an overview of the extension of the Security Abstration Model (SAM) and evaluates this security technology through expert interviews. The results demonstrate that our approach is compliant not only in the automotive domain, but also with standards for embedded systems from other domains. It thus provides a comprehensive and universal approach for the development of secure systems.}, language = {en} } @incollection{FischerTolvanenTavakoliKolagari2025, author = {Fischer, Alexander and Tolvanen, Juha-Pekka and Tavakoli Kolagari, Ramin}, title = {Embedded Systems Security Co-design: Modeling Support for Managers and Developers}, series = {Lecture Notes in Business Information Processing}, booktitle = {Lecture Notes in Business Information Processing}, publisher = {Springer Nature Switzerland}, address = {Cham}, isbn = {978-3-031-84913-8}, issn = {1865-1348}, doi = {10.1007/978-3-031-84913-8_8}, pages = {206 -- 232}, year = {2025}, abstract = {The proliferation of connected and autonomous vehicle technologies has significantly increased cybersecurity risks. Modern vehicles, as complex and networked computer systems, require comprehensive protection against malicious external attacks, much like conventional computers. Addressing these challenges requires robust tools that align established automotive model-based development approaches with the ISO/SAE 21434 standard for automotive cybersecurity, which became mandatory following its publication in 2021. Building on prior research, this paper introduces key innovations in the conceptual framework and the tool support that integrate seamlessly into existing automotive development methodologies. These advancements are rooted in extensions to the Security Abstraction Model (SAM) informed by the ISO/SAE 21434 standard. Notably, SAM now incorporates advanced methods for score calculation, including an attack potential-based approach for assessing attack feasibility and the computation of risk scores using risk matrices. Usability improvements are also a contribution, achieved through the introduction of BPMN-style (Business Process Model and Notation) diagrams tailored for the accessible visualization of otherwise complex security models. These diagrams make multifaceted attack trees easier to interpret, enabling managers and other non-technical stakeholders to intuitively understand security vulnerabilities and make informed decisions. Additionally, the tool supports updated metrics for impact and risk analysis, demonstrated through practical applications involving automotive subsystems such as braking. These examples illustrate improved traceability between SAM and functional design, ensuring that cybersecurity requirements are effectively integrated into the broader development lifecycle.}, language = {en} } @inproceedings{FischerBurkTavakoliKolagarietal.2025, author = {Fischer, Alexander and Burk, Louis and Tavakoli Kolagari, Ramin and Wienkop, Uwe}, title = {Machine-Readable by Design: Language Specifications as the Key to Integrating LLMs into Industrial Tools}, series = {Annals of Computer Science and Information Systems}, volume = {43}, booktitle = {Annals of Computer Science and Information Systems}, publisher = {IEEE}, isbn = {978-83-973291-6-4}, issn = {2300-5963}, doi = {10.15439/2025F5613}, pages = {531 -- 542}, year = {2025}, abstract = {We propose a meta-language-based approach enabling Large Language Models (LLMs) to reliably generate structured, machine-readable artifacts referred to as Meta-Language-defined Structures (MLDS) adapted to domain requirements, without adhering strictly to standard formats like JSON or XML. By embedding explicit schema instructions within prompts, we evaluated the method across diverse use cases, including automated Virtual Reality environment generation and automotive security modeling. Our experiments demonstrate that the meta-language approach significantly improves LLM-generated structure compliance, with an 88 \% validation rate across 132 test scenarios. Compared to traditional methods using LangChain and Pydantic, our MLDS method reduces setup complexity by approximately 80 \%, despite a marginally higher error rate. Furthermore, the MLDS artifacts produced were easily editable, enabling rapid iterative refinement. This flexibility greatly alleviates the "blank page syndrome" by providing structured initial artifacts suitable for immediate use or further human enhancement, making our approach highly practical for rapid prototyping and integration into complex industrial workflows.}, language = {en} } @inproceedings{BaumannWagnerRiedhammeretal.2025, author = {Baumann, Ilja and Wagner, Dominik and Riedhammer, Korbinian and Bocklet, Tobias}, title = {Pathology-Aware Speech Encoding and Data Augmentation for Dysarthric Speech Recognition}, series = {Interspeech 2025}, booktitle = {Interspeech 2025}, publisher = {ISCA}, address = {ISCA}, issn = {2958-1796}, doi = {10.21437/Interspeech.2025-2724}, pages = {3289 -- 3293}, year = {2025}, abstract = {Automatic speech recognition (ASR) for pathologic speech remains a major challenge due to high variability in articulation, phonation, and prosody distortions. In this work, we propose a pathology-aware speech encoder based on BEST-RQ pre-training, which incorporates 46k hours of speech, including pathologic and atypical speech. We continue pre-training for domain adaptation and experiment with etiology-specific codebooks. We achieve a 13.2\% relative word error rate (WER) improvement using the pathology-aware speech encoder with etiology-specific continued pre-training. Additionally, we examine the impact of incorporating synthetic and out-of-domain (OOD) data to further enhance ASR performance. Synthetic data reduces WER by up to 8.7\%, while OOD data improves WER by 12.2\%. Finally, we introduce a semantic similaritybased data augmentation technique to optimize data selection, achieving a WER improvement of up to 9.7\% while minimizing the need for additional training data.}, language = {en} } @article{RanzenbergerBaumannBayerletal.2025, author = {Ranzenberger, Thomas and Baumann, Ilja and Bayerl, Sebastian and Wagner, Dominik and Bocklet, Tobias and Riedhammer, Korbinian}, title = {Evaluation of recognition errors of hybrid and transformer-based ASR systems in German video lectures}, series = {Studientexte zur Sprachkommunikation: Elektronische Sprachsignalverarbeitung 2025 - Book}, journal = {Studientexte zur Sprachkommunikation: Elektronische Sprachsignalverarbeitung 2025 - Book}, publisher = {ESSV 2025}, address = {Halle, Deutschland}, pages = {101-108}, year = {2025}, abstract = {We analyze different errors in speech recognition systems, focusing on consecutive insertions and deletions, known as hallucinations and elisions in transformer-based end-to-end automatic speech recognition (ASR) systems. We compare errors from a TDNN-HMM, and whisper-based models on English and German spontaneous speech. Based on a human annotated subset of German lecture videos, we investigate whether these blocks of deletions affect the semantics of the utterance. Whisper performs best and preserves the meaning in 90\% of the annotated error segments even containing consecutive deletions on this subset. We analyze the word error rate and do further analysis of errors using natural language processing to detect lemmatization errors, compound word errors, and out-of-vocabulary words. We discuss possible reasons and mitigations.}, language = {en} } @inproceedings{BaumannWagnerRiedhammeretal.2025, author = {Baumann, Ilja and Wagner, Dominik and Riedhammer, Korbinian and Bocklet, Tobias}, title = {Optimized Self-supervised Training with BEST-RQ for Speech Recognition}, series = {ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, booktitle = {ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, publisher = {IEEE}, doi = {10.1109/ICASSP49660.2025.10889362}, pages = {1 -- 5}, year = {2025}, abstract = {Self-supervised learning has been successfully used for various speech related tasks, including automatic speech recognition. BERT-based Speech pre-Training with Random-projection Quantizer (BEST-RQ) has achieved state-of-the-art results in speech recognition. In this work, we further optimize the BEST-RQ approach using Kullback-Leibler divergence as an additional regularizing loss and multicodebook extension per cluster derived from low-level feature clustering. Preliminary experiments on train-100 split of LibriSpeech result in a relative improvement of 11.2\% on test-clean by using multiple codebooks, utilizing a combination of cross-entropy and Kullback-Leibler divergence further reduces the word error rate by 4.5\%. The proposed optimizations on full LibriSpeech pre-training and fine-tuning result in relative word error rate improvements of up to 23.8\% on test-clean and 30.6\% on testother using 6 codebooks. Furthermore, the proposed setup leads to faster convergence in pre-training and fine-tuning and additionally stabilizes the pre-training.}, language = {en} } @inproceedings{FreisingerSeebergerRanzenbergeretal.2025, author = {Freisinger, Steffen and Seeberger, Philipp and Ranzenberger, Thomas and Bocklet, Tobias and Riedhammer, Korbinian}, title = {Towards Multi-Level Transcript Segmentation: LoRA Fine-Tuning for Table-of-Contents Generation}, series = {Interspeech 2025}, booktitle = {Interspeech 2025}, publisher = {ISCA}, address = {ISCA}, issn = {2958-1796}, doi = {10.21437/Interspeech.2025-2792}, pages = {276 -- 280}, year = {2025}, abstract = {Segmenting speech transcripts into thematic sections benefits both downstream processing and users who depend on written text for accessibility. We introduce a novel approach to hierarchical topic segmentation in transcripts, generating multi-level tables of contents that capture both topic and subtopic boundaries. We compare zero-shot prompting and LoRA fine-tuning on large language models, while also exploring the integration of high-level speech pause features. Evaluations on English meeting recordings and multilingual lecture transcripts (Portuguese, German) show significant improvements over established topic segmentation baselines. Additionally, we adapt a common evaluation measure for multi-level segmentation, taking into account all hierarchical levels within one metric.}, language = {en} } @inproceedings{WagnerBaumannBocklet2025, author = {Wagner, Dominik and Baumann, Ilja and Bocklet, Tobias}, title = {Vocoder-Free Non-parallel Conversion of Whispered Speech With Masked Cycle-Consistent Generative Adversarial Networks}, publisher = {Springer}, address = {Cham}, isbn = {978-3-032-02548-7}, doi = {10.1007/978-3-032-02548-7_20}, pages = {235-246}, year = {2025}, abstract = {Cycle-consistent generative adversarial networks have been widely used in non-parallel voice conversion (VC). Their ability to learn mappings between source and target features without relying on parallel training data eliminates the need for temporal alignments. However, most methods decouple the conversion of acoustic features from synthesizing the audio signal by using separate models for conversion and waveform synthesis. This work unifies conversion and synthesis into a single model, thereby eliminating the need for a separate vocoder. By leveraging cycle-consistent training and a self-supervised auxiliary training task, our model is able to efficiently generate converted high-quality raw audio waveforms. Subjective listening tests showed that our unified approach achieved improvements of up to 6.7\% relative to the baseline in whispered VC. Mean opinion score predictions also yielded stable results in conventional VC (between 0.5\% and 2.4\% relative improvement).}, language = {en} } @inproceedings{WagnerBaumannEngertetal.2025, author = {Wagner, Dominik and Baumann, Ilja and Engert, Natalie and Lee, Seanie and N{\"o}th, Elmar and Riedhammer, Korbinian and Bocklet, Tobias}, title = {Personalized Fine-Tuning with Controllable Synthetic Speech from LLM-Generated Transcripts for Dysarthric Speech Recognition}, series = {Interspeech 2025}, booktitle = {Interspeech 2025}, publisher = {ISCA}, address = {ISCA}, issn = {2958-1796}, doi = {10.21437/Interspeech.2025-2155}, pages = {3294 -- 3298}, year = {2025}, abstract = {In this work, we present our submission to the Speech Accessibility Project challenge for dysarthric speech recognition. We integrate parameter-efficient fine-tuning with latent audio representations to improve an encoder-decoder ASR system. Synthetic training data is generated by fine-tuning Parler-TTS to mimic dysarthric speech, using LLM-generated prompts for corpus-consistent target transcripts. Personalization with x-vectors consistently reduces word error rates (WERs) over non-personalized fine-tuning. AdaLoRA adapters outperform full fine-tuning and standard low-rank adaptation, achieving relative WER reductions of ∼23\% and ∼22\%, respectively. Further improvements (∼5\% WER reduction) come from incorporating wav2vec 2.0-based audio representations. Training with synthetic dysarthric speech yields up to ∼7\% relative WER improvement over personalized fine-tuning alone.}, language = {en} } @inproceedings{LeeSeongBokLeeetal.2025, author = {Lee, Seanie and Seong, Haebin and Bok Lee, Dong and Kang, Minki and Chen, Xiaoyin and Wagner, Dominik and Bengio, Joshua and Lee, Juho and Hwang, Sung Ju}, title = {HarmAug: Effective Data Augmentation for Knowledge Distillation of Safety Guard Models}, publisher = {ICLR}, doi = {10.48550/arXiv.2410.01524}, pages = {1 -- 20}, year = {2025}, abstract = {Safety guard models that detect malicious queries aimed at large language models(LLMs) are essential for ensuring the secure and responsible deployment of LLMs in real-world applications. However, deploying existing safety guard models with billions of parameters alongside LLMs on mobile devices is impractical due to substantial memory requirements and latency. To reduce this cost, we distill a large teacher safety guard model into a smaller one using a labeled dataset of instruction-response pairs with binary harmfulness labels. Due to the limited diversity of harmful instructions in the existing labeled dataset, naively distilled models tend to underperform compared to larger models. To bridge the gap between small and large models, we propose HarmAug, a simple yet effective data augmentation method that involves jailbreaking an LLM and prompting it to generate harmful instructions. Given a prompt such as, "Make a single harmful instruction prompt that would elicit offensive content", we add an affirmative prefix (e.g., "I have an idea for a prompt:") to the LLM's response. This encourages the LLM to continue generating the rest of the response, leading to sampling harmful instructions. Another LLM generates a response to the harmful instruction, and the teacher model labels the instruction-response pair. We empirically show that our HarmAug outperforms other relevant baselines. Moreover, a 435-millionparameter safety guard model trained with HarmAug achieves an F1 score comparable to larger models with over 7 billion parameters, and even outperforms them in AUPRC, while operating at less than 25\% of their computational cost. Our code, safety guard model, and synthetic dataset are publicly available.}, language = {en} } @inproceedings{WagnerChurchillSigtiaetal.2025, author = {Wagner, Dominik and Churchill, Alexander and Sigtia, Siddharth and Marchi, Erik}, title = {SELMA: A Speech-Enabled Language Model for Virtual Assistant Interactions}, series = {ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, booktitle = {ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, publisher = {IEEE}, issn = {2379-190X}, doi = {10.1109/ICASSP49660.2025.10890139}, pages = {1 -- 5}, year = {2025}, abstract = {In this work, we present and evaluate SELMA, a Speech-Enabled Language Model for virtual Assistant interactions that integrates audio and text as inputs to a Large Language Model (LLM). SELMA is designed to handle three primary and two auxiliary tasks related to interactions with virtual assistants simultaneously within a single end-to-end model. We employ low-rank adaptation modules for parameter-efficient training of both the audio encoder and the LLM. Additionally, we implement a feature pooling strategy enabling the system to recognize global patterns and improve accuracy on tasks less reliant on individual sequence elements. Experimental results on Voice Trigger (VT) detection, Device-Directed Speech Detection (DDSD), and Automatic Speech Recognition (ASR), demonstrate that our approach both simplifies the typical input processing pipeline of virtual assistants significantly and also improves performance compared to dedicated models for each individual task. SELMA yields relative Equal-Error Rate improvements of 64\% on the VT detection task, and 22\% on DDSD, while also achieving word error rates close to the baseline.}, language = {en} } @inproceedings{BundschererSchmittBaumannetal.2025, author = {Bundscherer, Maximilian and Schmitt, Thomas and Baumann, Ilja and Bocklet, Tobias}, title = {Digital Operating Mode Classification of Real-World Amateur Radio Transmissions}, series = {ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, booktitle = {ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, publisher = {IEEE}, doi = {10.1109/ICASSP49660.2025.10889837}, pages = {1 -- 5}, year = {2025}, abstract = {This study presents an ML approach for classifying digital radio operating modes evaluated on real-world transmissions. We generated 98 different parameterized radio signals from 17 digital operating modes, transmitted each of them on the 70 cm (UHF) amateur radio band, and recorded our transmissions with two different architectures of SDR receivers. Three lightweight ML models were trained exclusively on spectrograms of limited non-transmitted signals with random characters as payloads. This training involved an online data augmentation pipeline to simulate various radio channel impairments. Our best model, EfficientNetB0, achieved an accuracy of 93.80\% across the 17 operating modes and 85.47\% across all 98 parameterized radio signals, evaluated on our real-world transmissions with Wikipedia articles as payloads. Furthermore, we analyzed the impact of varying signal durations \& the number of FFT bins on classification, assessed the effectiveness of our simulated channel impairments, and tested our models across multiple simulated SNRs.}, language = {en} } @masterthesis{Brandes2025, type = {Bachelor Thesis}, author = {Brandes, Joshua}, title = {Reactive Programming in Java: An Analysis of Application Performance and Developer Experience}, address = {N{\"u}rnberg}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:92-opus4-42529}, school = {Technische Hochschule N{\"u}rnberg Georg Simon Ohm}, pages = {112}, year = {2025}, abstract = {Diese Arbeit untersucht die praktischen Auswirkungen der Einf{\"u}hrung reaktiver Programmierung in Java-Anwendungen mit besonderem Augenmerk auf Performance und Entwickler*innenerfahrung. Durch die Implementierung und Analyse von Testanwendungen, die mit Project Reactor erstellt wurden, bewertet die Studie die Effektivit{\"a}t reaktiver Ans{\"a}tze bei unterschiedlichen Workloads, insbesondere im Kontext von Webanwendungen. Zus{\"a}tzlich bietet eine Umfrage unter Softwareentwickler*innen Einblicke in die Entwickler*innenzufriedenheit und typische Herausforderungen bei der Umsetzung. Dar{\"u}ber hinaus wird die Code-Komplexit{\"a}t reaktiver Programmierung gegen{\"u}ber klassischer imperativer Programmierung evaluiert. Alternative Frameworks wie RxJava sowie neuere Ans{\"a}tze wie Java Virtual Threads werden ebenfalls betrachtet. Die Ergebnisse zeigen, dass reaktiver Code von Entwickler*innen als komplexer wahrgenommen wird, obwohl klassische Metriken diese Komplexit{\"a}t oft untersch{\"a}tzen. Performanceverbesserungen sind vor allem bei I/O-gebundenen Workloads erkennbar, w{\"a}hrend CPU-intensive Aufgaben weniger profitieren. Die Ergebnisse erm{\"o}glichen Organisationen wie DATEV eG, fundierte Entscheidungen bez{\"u}glich der Einf{\"u}hrung reaktiver Programmierung zu treffen, indem Performancevorteile systematisch dem zus{\"a}tzlichen Entwicklungsaufwand gegen{\"u}bergestellt werden.}, language = {en} } @inproceedings{ElBardouhSchuhbauer2025, author = {El-Bardouh, Jasmin and Schuhbauer, Heidi}, title = {Auswirkungen des EU AI Act auf den Einsatz von K{\"u}nstlicher Intelligenz in der Hochschullehre}, series = {Tagungsband zum 6. Symposium zur Hochschullehre in den MINT-F{\"a}chern}, volume = {2025}, booktitle = {Tagungsband zum 6. Symposium zur Hochschullehre in den MINT-F{\"a}chern}, pages = {309-314}, year = {2025}, abstract = {Bei der Untersuchung der Auswirkungen des AI Act auf den Einsatz von KI in der Hochschullehre wird die Mehrheit der relevanten Tools der Kategorie des transparenzpflichtigen Risikos zugewiesen. Entscheidend f{\"u}r die daraus entstehenden Pflichten ist, ob die Hochschule Software selbst entwickelt und anbietet oder nur betreibt. Die Ermittlung der indirekten Auswirkungen erfolgt unter Zuhilfenahme des PESTEL-Modells. Zu den hochschulpolitischen Folgen z{\"a}hlt insbesondere die Notwendigkeit einer Anpassung der Lehrpl{\"a}ne sowie eine Weiterbildung des Hochschulpersonals. Auf soziokultureller Ebene l{\"a}sst sich eine verst{\"a}rkte Akzeptanz von KI-Systemen beobachten. Gleichzeitig birgt aus wirtschaftlicher Perspektive die Einf{\"u}hrung strengerer Anforderungen an Hochrisiko-KI-Systeme eine Kostensteigerung. Aus technologischer Perspektive l{\"a}sst sich eine zunehmende Relevanz neuer Tools im Hochschulbereich beobachten. Die rechtliche Dimension betont die Bedeutung von Compliance-Management-Systemen.}, language = {de} } @article{AltenberndGustafssonLisperetal.2016, author = {Altenbernd, Peter and Gustafsson, Jan and Lisper, Bj{\"o}rn and Stappert, Friedhelm}, title = {Early execution time-estimation through automatically generated timing models}, series = {Real-Time Systems}, volume = {52}, journal = {Real-Time Systems}, number = {6}, publisher = {Springer Science and Business Media LLC}, issn = {0922-6443}, doi = {10.1007/s11241-016-9250-7}, pages = {731 -- 760}, year = {2016}, abstract = {Traditional timing analysis, such as worst-case execution time analysis, is normally applied only in the late stages of embedded system software development, when the hardware is available and the code is compiled and linked. However, preliminary timing estimates are often needed in early stages of system development as an essential prerequisite for the configuration of the hardware setup and dimensioning of the system. During this phase the hardware is often not available, and the code might not be ready to link. This article describes an approach to predict the execution time of software through an early, source-level timing analysis. A timing model for source code is automatically derived from a given combination of hardware architecture and compiler. The model is identified from measured execution times for a set of synthetic training programs, compiled for the hardware platform in question. It can be used to estimate the execution time for code running on the platform: the estimation is then done directly from the source code, without compiling and running it. Our experiments show that, using this model, we can predict the execution times of the final, compiled code surprisingly well. For instance, we achieve an average deviation of 8\% for a set of benchmark programs for the ARM7 architecture.}, language = {en} } @inproceedings{RanzenbergerBockletFreisingeretal.2024, author = {Ranzenberger, Thomas and Bocklet, Tobias and Freisinger, Steffen and Georges, Munir and Glockner, Kevin and Herygers, Aaricia and Riedhammer, Korbinian and Schneider, Fabian and Simic, Christopher and Zakaria, Khabbab}, title = {EXTENDING HANS: LARGE LANGUAGE MODELS FOR QUESTION ANSWERING, SUMMARIZATION, AND TOPIC SEGMENTATION IN AN ML-BASED LEARNING EXPERIENCE PLATFORM}, series = {Elektronische Sprachsignalverarbeitung 2024, Tagungsband der 35. Konferenz, Regensburg, 6.-8. M{\"a}rz 2024}, booktitle = {Elektronische Sprachsignalverarbeitung 2024, Tagungsband der 35. Konferenz, Regensburg, 6.-8. M{\"a}rz 2024}, publisher = {TUPress}, address = {Dresden}, isbn = {978-3-95908-325-6}, pages = {219-224}, year = {2024}, abstract = {Abstract: The use of chatbots based on large language models (LLMs) and their impact on society are influencing our learning experience platform Hochschul Assistenz-System (HAnS). HAnS uses machine learning (ML) methods to support students and lecturers in the online learning and teaching processes [1]. This paper introduces LLM-based features available in HAnS which are using the transcript of our improved Automatic Speech Recognition (ASR) pipeline with an average transcription duration of 45 seconds and an average word error rate (WER) of 6.66\% on over 8 hours of audio data of 7 lecture videos. A LLM-based chatbot could be used to answer questions on the lecture content as the ASR transcript is provided as context. The summarization and topic segmentation uses the LLM to improve our learning experience platform. We generate multiple choice questions using the LLM and the ASR transcript as context during playback in a period of 3 minutes and display them in the HAnS frontend}, language = {en} } @inproceedings{WagnerLeeBaumannetal.2024, author = {Wagner, Dominik and Lee, Seanie and Baumann, Ilja and Seeberger, Philipp and Riedhammer, Korbinian and Bocklet, Tobias}, title = {Optimized Speculative Sampling for {GPU} Hardware Accelerators}, series = {Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing}, booktitle = {Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing}, publisher = {Association for Computational Linguistics}, address = {Miami, Florida, USA}, doi = {10.18653/v1/2024.emnlp-main.370}, pages = {6442-6458}, year = {2024}, abstract = {In this work, we optimize speculative sampling for parallel hardware accelerators to improve sampling speed. We notice that substantial portions of the intermediate matrices necessary for speculative sampling can be computed concurrently. This allows us to distribute the workload across multiple GPU threads, enabling simultaneous operations on matrix segments within thread blocks. This results in profiling time improvements ranging from 6\% to 13\% relative to the baseline implementation, without compromising accuracy. To further accelerate speculative sampling, probability distributions parameterized by softmax are approximated by sigmoid. This approximation approach results in significantly greater relative improvements in profiling time, ranging from 37\% to 94\%, with a minor decline in accuracy. We conduct extensive experiments on both automatic speech recognition and summarization tasks to validate the effectiveness of our optimization methods.}, language = {en} } @article{WagnerBaumannBocklet2024, author = {Wagner, Dominik and Baumann, Ilja and Bocklet, Tobias}, title = {Generative adversarial networks for whispered to voiced speech conversion: a comparative study}, series = {International Journal of Speech Technology}, volume = {27}, journal = {International Journal of Speech Technology}, editor = {Baumann, Ilja}, doi = {10.1007/s10772-024-10161-1}, pages = {1093-1110}, year = {2024}, abstract = {Generative Adversarial Networks (GANs) have demonstrated promising results as end-to-end models for whispered to voiced speech conversion. Leveraging non-autoregressive systems like GANs capable of performing conditional waveform generation eliminates the need for separate models to estimate voiced speech features, and leads to faster inference compared to autoregressive methods. This study aims to identify the optimal GAN architecture for the whispered to voiced speech conversion task by comparing six state-of-the-art models. Furthermore, we present a method for evaluating the preservation of speaker identity and local accent, using embeddings obtained from speaker- and language identification systems. Our experimental results show that building the speech conversion system based on the HiFi-GAN architecture yields the best objective evaluation scores, outperforming the baseline by ∼9\% relative using frequency-weighted Signal-to-Noise Ratio and Log Likelihood Ratio, as well as by ∼29\% relative using Root Mean Squared Error. In subjective tests, HiFi-GAN yielded a mean opinion score of 2.9, significantly outperforming the baseline with a score of 1.4. Furthermore, HiFi-GAN enhanced ASR performance and preserved speaker identity and accent, with correct language detection rates of up to ∼98\%.}, language = {en} } @inproceedings{RanzenbergerFreierReinoldetal.2024, author = {Ranzenberger, Thomas and Freier, Carolin and Reinold, Luca and Riedhammer, Korbinian and Schneider, Fabian and Simic, Christopher and Simon, Claudia and Freisinger, Steffen and Georges, Munir and Bocklet, Tobias}, title = {A Multidisciplinary Approach to AI-based self-motivated Learning and Teaching with Large Language Models}, series = {Proceedings of DELFI Workshops 2024}, booktitle = {Proceedings of DELFI Workshops 2024}, publisher = {Gesellschaft f{\"u}r Informatik e.V.}, doi = {10.18420/delfi2024_11}, year = {2024}, abstract = {We present a learning experience platform that uses machine learning methods to support students and lecturers in self-motivated online learning and teaching processes. The platform is being developed as an agile open-source collaborative project supported by multiple universities and partners. The development is guided didactically, reviewed, and scientifically evaluated in several cycles. Transparency, data protection and the copyright compliant use of the system is a central part of the project. The system further employs large language models (LLMs). Due to privacy concerns, we utilize locally hosted LLM instances and explicitly do not rely on available cloud products. Students and lecturers can interact with an LLM-based chatbot in the current prototype. The AI-generated outputs contain cross-references to the current educational video's context, indicating if sections are based on the lectures context or world knowledge. We present the prototype and results of our qualitative evaluation from the perspective of lecturers and students.}, language = {en} } @inproceedings{SeebergerWagnerRiedhammer2024, author = {Seeberger, Philipp and Wagner, Dominik and Riedhammer, Korbinian}, title = {Multimodal Multimedia Event Argument Extraction with Unified Template Filling}, series = {Findings of the Association for Computational Linguistics: EMNLP 2024}, booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2024}, publisher = {Association for Computational Linguistics}, address = {Miami, Florida, USA}, doi = {10.18653/v1/2024.findings-emnlp.381}, pages = {6539-6548}, year = {2024}, abstract = {With the advancement of multimedia technologies, news documents and user-generated content are often represented as multiple modalities, making Multimedia Event Extraction (MEE) an increasingly important challenge. However, recent MEE methods employ weak alignment strategies and data augmentation with simple classification models, which ignore the capabilities of natural language-formulated event templates for the challenging Event Argument Extraction (EAE) task. In this work, we focus on EAE and address this issue by introducing a unified template filling model that connects the textual and visual modalities via textual prompts. This approach enables the exploitation of cross-ontology transfer and the incorporation of event-specific semantics. Experiments on the M2E2 benchmark demonstrate the effectiveness of our approach. Our system surpasses the current SOTA on textual EAE by +7\% F1, and performs generally better than the second-best systems for multimedia EAE.}, language = {en} } @inproceedings{WagnerBaumannRanzenbergeretal.2024, author = {Wagner, Dominik and Baumann, Ilja and Ranzenberger, Thomas and Riedhammer, Korbinian and Bocklet, Tobias}, title = {Personalizing Large Sequence-to-Sequence Speech Foundation Models With Speaker Representations}, doi = {10.1109/SLT61566.2024.10832252}, pages = {1-6}, year = {2024}, abstract = {We present a method to personalize large transformer-based encoderdecoder speech foundation models without the need for changes in the underlying model structure or training from scratch. This is achieved by projecting speaker-specific information into the latent space of the transformer decoder via a small neural network and learning to process the speaker information along with domainspecific information via parameter-efficient finetuning. We use this method to improve the automatic speech recognition results of spoken academic German and English. Our approach yields average relative word error rate (WER) improvements of approximately 29\% on German academic speech and 25\% on English academic speech. It also translates well to conversational speech, achieving relative WER improvements of up to 36\%, and demonstrates modest gains of up to 5\% on read speech. Moreover, we observe that incorporating utterances from the recent past as personalization context yields the most significant overall improvements and that changes in voice characteristics resulting from prolonged speaking have a minimal effect on the personalization quality of academic lectures.}, language = {en} } @inproceedings{BraunBayerlHoenigetal.2024, author = {Braun, Franziska and Bayerl, Sebastian and H{\"o}nig, Florian and Lehfeld, Hartmut and Hillemacher, Thomas and Bocklet, Tobias and Riedhammer, Korbinian}, title = {Infusing Acoustic Pause Context into Text-Based Dementia Assessment}, issn = {2958-1796}, doi = {10.21437/Interspeech.2024-2496}, pages = {1980-1984}, year = {2024}, abstract = {Speech pauses, alongside content and structure, offer a valuable and non-invasive biomarker for detecting dementia. This work investigates the use of pause-enriched transcripts in transformer-based language models to differentiate the cognitive states of subjects with no cognitive impairment, mild cognitive impairment, and Alzheimer's dementia based on their speech from a clinical assessment. We address three binary classification tasks: Onset, monitoring, and dementia exclusion. The performance is evaluated through experiments on a German Verbal Fluency Test and a Picture Description Test, comparing the model's effectiveness across different speech production contexts. Starting from a textual baseline, we investigate the effect of incorporation of pause information and acoustic context. We show the test should be chosen depending on the task, and similarly, lexical pause information and acoustic cross-attention contribute differently.}, language = {en} } @inproceedings{DomanskiReifSchaabetal.2025, author = {Domanski, Jan and Reif, Jonas and Schaab, Lea and Sch{\"o}tteler, Sebastian and Schuhbauer, Heidi}, title = {JUDGED BY ITS COVER: UNDERSTANDING THE FORMATION OF BOOK TESTIMONIAL NETWORKS}, series = {Proceedings of FEB Zagreb 16th International Odyssey Conference on Economics and Business}, booktitle = {Proceedings of FEB Zagreb 16th International Odyssey Conference on Economics and Business}, editor = {Sever Mališ, and Načinović Braje, and Raguž Krištić,}, publisher = {University of Zagreb, Faculty of Economics \& Business}, address = {Zagreb}, pages = {732-744}, year = {2025}, abstract = {Testimonials, short promotional statements often written by notable authors or critics, are a crucial marketing tool in the publishing industry, contributing to a book's market success. This study investigates the formation of ties in testimonial networks, focusing on both network-endogenous factors (reciprocity, transitivity and preferential attachment) and exogenous factors tied to social homophily, such as shared publisher, genre and gender. A network of 575 authors was constructed from testimonials in fiction books published between 2020 and 2024 and analyzed using Exponential Random Graph Models (ERGM). The results reveal strong effects of reciprocity, transitivity, and homophily, particularly with respect to publisher and genre. Reciprocity appears to play a key role in testimonial exchanges, suggesting that authors who provide testimonials are significantly more likely to receive testimonials in return. Authors are likely to form connections through shared relationships. However, the network does not exhibit significant inequality in tie distribution, indicating that preferential attachment is not a dominant mechanism. The results strongly confirm that shared attributes, especially publisher and genre, significantly influence testimonial formation, with publisher-based homophily being the most impactful. This reflects the role of publishers in orchestrating testimonial exchanges, as they may actively encourage authors within their networks to provide testimonials for one another. These findings offer insights into the structural dynamics of testimonial networks and provide practical implications for emerging authors seeking to establish connections in the competitive publishing market. The strong evidence for reciprocity underscores the im- portance of actively engaging in the testimonial exchange process. The lack of evidence for preferential attachment is a promising finding for emerging authors, as it suggests they are not disadvantaged by having fewer initial connections.}, language = {en} } @book{AxenieBauerLopezCoronaetal.2025, author = {Axenie, Cristian and Bauer, Roman and L{\´o}pez Corona, Oliver and West, Jeffrey}, title = {Applied Antifragility in Natural Systems}, publisher = {Springer Nature Switzerland}, address = {Cham}, isbn = {9783031903908}, issn = {2191-5768}, doi = {10.1007/978-3-031-90391-5}, pages = {XVI, 80}, year = {2025}, abstract = {As coined in the book of Nassim Taleb, antifragility is a property of a system to gain from uncertainty, randomness, and volatility, opposite to what fragility would incur. An antifragile system's response to external perturbations is beyond robust, such that small stressors can strengthen the future response of the system by adding a strong anticipation component. Such principles are already well suited for describing behaviors in natural systems but also in approaching therapy designs and eco-system modelling and eco-system analysis. The purpose of this book is to build a foundational knowledge base by applying antifragile system design, analysis, and development in natural systems, including biomedicine, neuroscience, and ecology as main fields. We are interested in formalizing principles and an apparatus that turns the basic concept of antifragility into a tool for designing and building closed-loop systems that behave beyond robust in the face of uncertainty when characterizing and intervening in biomedical and ecological (eco)systems. The book introduces the framework of applied antifragility and possible paths to build systems that gain from uncertainty. We draw from the body of literature on natural systems (e.g. cancer therapy, antibiotics, neuroscience, and agricultural pest management) in an attempt to unify the scales of antifragility in one framework. The work of the Applied Antifragility Group in oncology, neuroscience, and ecology led by the authors provides a good overview on the current research status.}, language = {en} } @book{AxenieAkbarzadehMakridisetal.2025, author = {Axenie, Cristian and Akbarzadeh, Meisam and Makridis, Michail A. and Saveriano, Matteo and Stancu, Alexandru}, title = {Applied Antifragility in Technical Systems}, publisher = {Springer Nature Switzerland}, address = {Cham}, isbn = {9783031904240}, issn = {2191-5768}, doi = {10.1007/978-3-031-90425-7}, pages = {XVI, 122}, year = {2025}, abstract = {The book purpose is to build a foundational knowledge base by applying antifragile system design, analysis, and development in technical systems, with a focus on traffic engineering, robotics, and control engineering. The authors are interested in formalizing principles and an apparatus that turns the basic concept of antifragility into a tool for designing and building closed-loop technical systems that behave beyond robust in the face of uncertainty. As coined in the book of Nassim Taleb, antifragility is a property of a system to gain from uncertainty, randomness, and volatility, opposite to what fragility would incur. An antifragile system's response to external perturbations is beyond robust, such that small stressors can strengthen the future response of the system by adding a strong anticipation component. The work of the Applied Antifragility Group in traffic control and robotics, led by the authors, provides a good overview on the current research status.}, language = {en} } @inproceedings{ArzbergerTavakoliKolagari2025, author = {Arzberger, Alexandra and Tavakoli Kolagari, Ramin}, title = {Hi-ALPS - An Experimental Robustness Quantification of Six LiDAR-based Object Detection Systems for Autonomous Driving}, series = {2025 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)}, booktitle = {2025 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)}, publisher = {IEEE}, doi = {10.1109/SaTML64287.2025.00050}, pages = {811 -- 823}, year = {2025}, abstract = {Light Detection and Ranging (LiDAR) is an essential sensor technology for autonomous driving as it can capture high-resolution 3D data. As 3D object detection systems (OD) can interpret such point cloud data, they play a key role in the driving decisions of autonomous vehicles. Consequently, such 3D OD must be robust against all types of perturbations and must therefore be extensively tested. One approach is the use of adversarial examples, which are small, sometimes sophisticated perturbations in the input data that change, i.e., falsify, the prediction of the OD. These perturbations are carefully designed based on the weaknesses of the OD. The robustness of the OD cannot be quantified with adversarial examples in general, because if the OD is vulnerable to a given attack, it is unclear whether this is due to the robustness of the OD or whether the attack algorithm produces particularly strong adversarial examples. The contribution of this work is Hi-ALPS -- Hierarchical Adversarial-example-based LiDAR Perturbation Level System, where higher robustness of the OD is required to withstand the perturbations as the perturbation levels increase. In doing so, the Hi-ALPS levels successively implement a heuristic followed by established adversarial example approaches. In a series of comprehensive experiments using Hi-ALPS, we quantify the robustness of six state-of-the-art 3D OD under different types of perturbations. The results of the experiments show that none of the OD is robust against all Hi-ALPS levels; an important factor for the ranking is that human observers can still correctly recognize the perturbed objects, as the respective perturbations are small. To increase the robustness of the OD, we discuss the applicability of state-of-the-art countermeasures. In addition, we derive further suggestions for countermeasures based on our experimental results.}, language = {en} } @inproceedings{vonRymonLipinskiSeibtRothetal.2019, author = {von Rymon Lipinski, Bartosz and Seibt, Simon and Roth, Johannes and Ab{\´e}, Dominik}, title = {Level Graph - Incremental Procedural Generation of Indoor Levels using Minimum Spanning Trees}, doi = {10.1109/CIG.2019.8847956}, pages = {1-7}, year = {2019}, abstract = {Procedural generation of virtual worlds is an important aspect of game development since decades, typically for increasing replayability or for speeding up the level creation process. However, the utilization of this potential has always been a great challenge due to the difficult controllability of the underlying algorithms or limitations to specific level geometries, like 2D regular structures. In this paper, we present a novel approach for semi-automatic generation of a wide variety of 2D/3D corridor and room systems. The underlying processing pipeline is based on a separation between a user-guided generation of a graph-based abstract level structure and a fully-automatic construction of the corresponding geometry using a pre-modeled component library. The core algorithm is built on the computation of an extended minimal spanning tree, which can be controlled by a set of intuitive vertex and edge parameters. First experimental results have shown that our incremental generation pipeline allows the efficient creation of complex indoor levels, minimizing limitations on level and game designers' creativity.}, language = {en} } @misc{SeibtChangvonRymonLipinskietal.2024, author = {Seibt, Simon and Chang, Thomas and von Rymon Lipinski, Bartosz and Latoschik, Marc Erich}, title = {Dense 3D Gaussian Splatting Initialization for Sparse Image Data}, doi = {10.2312/egp.20241038}, pages = {2}, year = {2024}, abstract = {This paper presents advancements in novel-view synthesis with 3D Gaussian Splatting (3DGS) using a dense and accurate SfM point cloud initialization approach. We address the challenge of achieving photorealistic renderings from sparse image data, where basic 3DGS training may result in suboptimal convergence, thus leading to visual artifacts. The proposed method enhances precision and density of initially reconstructed point clouds by refining 3D positions and extrapolating additional points, even for difficult image regions, e.g. with repeating patterns and suboptimal visual coverage. Our contributions focus on improving "Dense Feature Matching for Structure-from-Motion" (DFM4SfM) based on a homographic decomposition of the image space to support 3DGS training: First, a grid-based feature detection method is introduced for DFM4SfM to ensure a welldistributed 3D Gaussian initialization uniformly over all depth planes. Second, the SfM feature matching is complemented by a geometric plausibility check, priming the homography estimation and thereby improving the initial placement of 3D Gaussians. Experimental results on the NeRF-LLFF dataset demonstrate that this approach achieves superior qualitative and quantitative results, even for fewer views, and the potential for a significantly accelerated 3DGS training with faster convergence.}, language = {en} } @inproceedings{SeibtAroldvonRymonLipinskietal.2024, author = {Seibt, Simon and Arold, Michael and von Rymon Lipinski, Bartosz and Wienkopf, Uwe and Latoschik, Marc Erich}, title = {Parallax-Aware Image Stitching Based on Homographic Decomposition}, publisher = {Springer}, address = {Cham}, isbn = {978-3-031-54605-1}, doi = {10.1007/978-3-031-54605-1_13}, year = {2024}, abstract = {Image stitching plays a crucial role for various computer vision applications, like panoramic photography, video production, medical imaging and satellite imagery. It makes it possible to align two images captured at different views onto a single image with a wider field of view. However, for 3D scenes with high depth complexity and images captured from two different positions, the resulting image pair may exhibit significant parallaxes. Stitching images with multiple or large apparent motion shifts remains a challenging task, and existing methods often fail in such cases. In this paper, a novel image stitching pipeline is introduced, addressing the aforementioned challenge: First, iterative dense feature matching is performed, which results in a multi-homography decomposition. Then, this output is used to compute a per-pixel multidimensional weight map of the estimated homographies for image alignment via weighted warping. Additionally, the homographic image space decomposition is exploited using combinatorial analysis to identify parallaxes, resulting in a parallax-aware overlapping region: Parallax-free overlapping areas only require weighted warping and blending. For parallax areas, these operations are omitted to avoid ghosting artifacts. Instead, histogram- and mask-based color mapping is performed to ensure visual color consistency. The presented experiments demonstrate that the proposed method provides superior results regarding precision and handling of parallaxes.}, language = {en} } @inproceedings{ChangSeibtvonRymonLipinski2024, author = {Chang, Thomas and Seibt, Simon and von Rymon Lipinski, Bartosz}, title = {Hierarchical Histogram Threshold Segmentation - Auto-terminating High-detail Oversegmentation}, series = {2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, booktitle = {2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, doi = {10.1109/CVPR52733.2024.00308}, pages = {3195 -- 3204}, year = {2024}, abstract = {Superpixels play a crucial role in image processing by partitioning an image into clusters of pixels with similar visual attributes. This facilitates subsequent image processing tasks, offering computational advantages over the manipulation of individual pixels. While numerous oversegmentation techniques have emerged in recent years, many rely on predefined initialization and termination criteria. In this paper, a novel top-down superpixel segmentation algorithm called Hierarchical Histogram Threshold Segmentation (HHTS) is introduced. It eliminates the need for initialization and implements auto-termination, outperforming state-of-the-art methods w.r.t. boundary recall. This is achieved by iteratively partitioning individual pixel segments into foreground and background and applying intensity thresholding across multiple color channels. The underlying iterative process constructs a superpixel hierarchy that adapts to local detail distributions until color information exhaustion. Experimental results demonstrate the superiority of the proposed approach in terms of boundary adherence, while maintaining competitive runtime performance on the BSDS500 and NYUV2 datasets. Furthermore, an application of HHTS in refining machine learningbased semantic segmentation masks produced by the Segment Anything Foundation Model (SAM) is presented.}, language = {en} } @inproceedings{SeibtvonRymonLipinskiChangetal.2023, author = {Seibt, Simon and von Rymon Lipinski, Bartosz and Chang, Thomas and Latoschik, Marc Erich}, title = {DFM4SFM - Dense Feature Matching for Structure from Motion}, series = {2023 IEEE International Conference on Image Processing Challenges and Workshops (ICIPCW)}, booktitle = {2023 IEEE International Conference on Image Processing Challenges and Workshops (ICIPCW)}, doi = {10.1109/ICIPC59416.2023.10328368}, pages = {3678 -- 3682}, year = {2023}, abstract = {Structure from motion (SfM) is a fundamental task in computer vision and allows recovering the 3D structure of a stationary scene from an image set. Finding robust and accurate feature matches plays a crucial role in the early stages of SfM. So in this work, we propose a novel method for computing image correspondences based on dense feature matching (DFM) using homographic decomposition: The underlying pipeline provides refinement of existing matches through iterative rematching, detection of occlusions and extrapolation of additional matches in critical image areas between image pairs. Our main contributions are improvements of DFM specifically for SfM, resulting in global refinement and global extrapolation of image correspondences between related views. Furthermore, we propose an iterative version of the Delaunay-triangulation-based outlier detection algorithm for robust processing of repeated image patterns. Through experiments, we demonstrate that the proposed method significantlv improves the reconstruction accuracy.}, language = {en} } @inproceedings{Brockmann2024, author = {Brockmann, Patricia}, title = {Problem-Based E-Learning to Increase Motivationn of STEM-Students}, series = {Proceedings of the International Teaching and Education Conference}, volume = {1}, booktitle = {Proceedings of the International Teaching and Education Conference}, number = {1}, publisher = {Diamond Scientific Publishing}, issn = {3030-1521}, doi = {10.33422/teduconf.v1i1.746}, pages = {14}, year = {2024}, abstract = {First semester students often face major challenges in adjusting to university studies. Inexperienced students may not appreciate the practical relevance of abstract, theoretical concepts taught using traditional, instructor-centered lectures. Furthermore, the rise of generative artificial intelligence (GenAI) can mislead some to believe they no longer need to exert effort to learn academic skills. To encourage students to engage in learning material and thus improve retention rates, two complementary teaching methods were integrated into a large, introductory course for first semester STEM students: Problem-Based Learning (PBL) and gamification. Problem-Based Learning was implemented to make the subject matter more meaningful by simulating a real world experience: How to start up a small business. This entrepreneurship task was simulated in an online game, to increase student motivation. Quantitative questionnaires of student motivation and experiences with e-learning were evaluated. Based on input obtained in expert interviews, a fictitious startup enterprise was designed to serve as the used case. An online e-learning game was developed, which leveraged gamification elements to try to increase motivation. Each phase of the startup process was represented as one level of the e-learning game. The e-learning game was tested by a group of first semester students. Their opinions were collected using an anonymous online survey. Aggregated results of the survey are discussed and plans for further research are presented.}, language = {en} } @inproceedings{MarutschkeBrockmann2024, author = {Marutschke, Daniel Moritz and Brockmann, Patricia}, title = {A Conceptual Framework for Multidisciplinary Data Science Education Based on IoT Projects}, series = {International Workshop on Artificial Intelligence and Education (WAIE)}, volume = {2024}, booktitle = {International Workshop on Artificial Intelligence and Education (WAIE)}, publisher = {IEEE Xplore}, doi = {10.1109/WAIE63876.2024.00065}, pages = {331 -- 335}, year = {2024}, abstract = {The necessity for professionals to make predictions using digital data has been recognized for many decades. With its roots in statistical analysis and later implemented as data analytics, Data Science emerged as a widely accepted discipline. The data science life cycle begins by setting a goal and ends with reporting/reflecting on the gained results. It incorporates known analytical techniques, including modern artificial intelligence and machine learning techniques. While often considered an independent discipline, both engineering and liberal arts colleges have recognized the need for an understanding of data science in the general workforce. The literature still suggests a disconnect between the study fields and how the Internet of Things could be leveraged in Project-Based Learning (PBL) style data science education. In this paper, the authors propose a conceptual framework to teach multidisciplinary data science at university level to engineering and liberal arts students using data from real-world Internet of Things (IoT) projects.}, language = {en} } @inproceedings{MarutschkeBrockmann2024, author = {Marutschke, Daniel Moritz and Brockmann, Patricia}, title = {Multicultural Observations in a Three-University Global Distributed Software Development Course}, series = {Proceedings of the 2024 The 6th World Symposium on Software Engineering (WSSE)}, volume = {2024}, booktitle = {Proceedings of the 2024 The 6th World Symposium on Software Engineering (WSSE)}, publisher = {ACM Digital Library}, isbn = {979-8-4007-1708-6}, doi = {10.1145/3698062.3698077}, pages = {115 -- 120}, year = {2024}, abstract = {In our increasingly interconnected world, software engineering students need increased exposure to not only state of the art technology, but to its usage and development in multicultural environments. This paper reports on findings from a global software engineering course taught at three universities in Japan, Germany, and Indonesia. Questionnaire and interview data gathered ex-ante (before) and ex-post (after) the semester about cultural dimensions and UML usage are analyzed qualitatively and quantitatively. The perceived importance of cultural dimensions was compared to previous two-university classes. Experience reports from instructors are described. Unique dynamics in students' perception of cultural dimensions could be observed. UML challenges self-reported are consistent with previous research findings. An ex-ante to ex-post comparison supports the implementation of real-world projects taught in PBL-style. These findings together are presented as best practices and lessons learned.}, language = {en} } @inproceedings{SeibtChangLatoschiketal.2024, author = {Seibt, Simon and Chang, Thomas and Latoschik, Marc Erich and Kuth, Bastian and von Rymon Lipinski, Bartosz}, title = {Multidimensional image morphing-fast image-based rendering of open 3D and VR environments}, series = {Virtual Reality \& Intelligent Hardware}, volume = {7}, booktitle = {Virtual Reality \& Intelligent Hardware}, number = {2}, doi = {10.1016/j.vrih.2023.06.007}, pages = {155-172}, year = {2024}, abstract = {In recent years, the demand for interactive photorealistic three-dimensional (3D) environments has increased in various fields, including architecture, engineering, and entertainment. However, achieving a balance between the quality and efficiency of high-performance 3D applications and virtual reality (VR) remains challenging. Methods This study addresses this issue by revisiting and extending view interpolation for image-based rendering (IBR), which enables the exploration of spacious open environments in 3D and VR. Therefore, we introduce multimorphing, a novel rendering method based on the spatial data structure of 2D image patches, called the image graph. Using this approach, novel views can be rendered with up to six degrees of freedom using only a sparse set of views. The rendering process does not require 3D reconstruction of the geometry or per-pixel depth information, and all relevant data for the output are extracted from the local morphing cells of the image graph. The detection of parallax image regions during preprocessing reduces rendering artifacts by extrapolating image patches from adjacent cells in real-time. In addition, a GPU-based solution was presented to resolve exposure inconsistencies within a dataset, enabling seamless transitions of brightness when moving between areas with varying light intensities. Results Experiments on multiple real-world and synthetic scenes demonstrate that the presented method achieves high "VR-compatible" frame rates, even on mid-range and legacy hardware, respectively. While achieving adequate visual quality even for sparse datasets, it outperforms other IBR and current neural rendering approaches. Conclusions Using the correspondence-based decomposition of input images into morphing cells of 2D image patches, multidimensional image morphing provides high-performance novel view generation, supporting open 3D and VR environments. Nevertheless, the handling of morphing artifacts in the parallax image regions remains a topic for future research.}, language = {en} } @article{LatoschikVONRYMONLIPINSKISeibt2022, author = {Latoschik, Marc Erich and VON RYMON LIPINSKI, BARTOSZ and Seibt, Simon}, title = {Dense Feature Matching Based on Homographic Decomposition}, series = {IEEE Access}, volume = {10}, journal = {IEEE Access}, doi = {10.1109/ACCESS.2022.3152539}, pages = {99129-99149}, year = {2022}, abstract = {Finding robust and accurate feature matches is a fundamental problem in computer vision. However, incorrect correspondences and suboptimal matching accuracies lead to significant challenges for many real-world applications. In conventional feature matching, corresponding features in an image pair are greedily searched using their descriptor distance. The resulting matching set is then typically used as input for geometric model fitting methods to find an appropriate fundamental matrix and filter out incorrect matches. Unfortunately, this basic approach cannot solve all practical problems, such as fundamental matrix degeneration, matching ambiguities caused by repeated patterns and rejection of initially mismatched features without further reconsideration. In this paper we introduce a novel matching pipeline, which addresses all of the aforementioned challenges at once: First, we perform iterative rematching to give mismatched feature points a further chance for being considered in later processing steps. Thereby, we are searching for inliers that exhibit the same homographic transformation per iteration. The resulting homographic decomposition is used for refining matches, occlusion detection (e.g. due to parallaxes) and extrapolation of additional features in critical image areas. Furthermore, Delaunay triangulation of the matching set is utilized to minimize the repeated pattern problem and to implement focused matching. Doing so, enables us to further increase matching quality by concentrating on local image areas, defined by the triangular mesh. We present and discuss experimental results with multiple real-world matching datasets. Our contributions, besides improving matching recall and precision for image processing applications in general, also relate to use cases in image-based computer graphics.}, language = {en} } @inproceedings{BrockmannSchuhbauerBienert2025, author = {Brockmann, Patricia and Schuhbauer, Heidi and Bienert, Margot}, title = {Based E-Learning to Increase Motivation of Stem-Students : Gamification of A Startup Enterpriser}, series = {Proceedings of the International Teaching and Education Conference}, booktitle = {Proceedings of the International Teaching and Education Conference}, year = {2025}, abstract = {First semester students often face major challenges in adjusting to university studies. Inexperienced students may not appreciate the practical relevance of abstract, theoretical concepts taught using traditional, instructor-centered lectures. Furthermore, the rise of generative artificial intelligence (GenAI) can mislead some to believe they no longer need to exert effort to learn academic skills. To encourage students to engage in learning material and thus improve retention rates, two complementary teaching methods were integrated into a large, introductory course for first semester STEM students: Problem-Based Learning (PBL) and gamification. Problem-Based Learning was implemented to make the subject matter more meaningful by simulating a real-world experience: How to start up a small business. This entrepreneurship task was simulated in an online game, to increase student motivation. Quantitative questionnaires of student motivation and experiences with e-learning were evaluated. Based on input obtained in expert interviews, a fictitious startup enterprise was designed to serve as the used case. An online e-learning game was developed, which leveraged gamification elements to try to increase motivation. Each phase of the startup process was represented as one level of the e-learning game. The e-learning game was tested by a group of first semester students. Their opinions were collected using an anonymous online survey. Aggregated results of the survey are discussed and plans for further research are presented.}, language = {en} } @inproceedings{SchuhbauerBrockmann2025, author = {Schuhbauer, Heidi and Brockmann, Patricia}, title = {Exploring the Impact of a Digital Escape Room in Moodle on Students´ Motivation}, series = {Proceedings of the International Teaching and Education Conference}, volume = {1}, booktitle = {Proceedings of the International Teaching and Education Conference}, number = {1}, issn = {3030-1521}, doi = {10.33422/teduconf.v1i1.685}, pages = {1 - 14}, year = {2025}, abstract = {Knowledge Management is a core subject in the bachelor's program in Information Systems at the Nuremberg Institute of Technology. To enhance learning, a digital escape room was developed and implemented on the Moodle e-learning platform. The project addressed three research questions: Is Moodle a recommendable platform to implement a digital escape room? Are digital escape rooms a reasonable concept to teach knowledge? Are students motivated to gain knowledge with an escape room? A prototype was initially evaluated, revealing improved learning outcomes, particularly among students who enjoy quizzes. However, the quiz-based format of Moodle proved unsuitable for certain topics, and the platform's limited capabilities for game development posed challenges. The original escape room's duration of two to three hours was considered excessive by students, leading to a revised version with a more manageable length of 90 minutes. Usability testing, employing the think-aloud method, was conducted to refine the design before re-release. A subsequent pilot revealed that 70\% of students valued the escape room for knowledge acquisition. While many enjoyed the puzzles, some preferred traditional exam preparation. These findings suggest that digital escape rooms hold promise as innovative pedagogical tools. However, their success depends on careful adaptation to course content and student needs, making them a potential complement to traditional learning methods. The overarching goal of this project is to evaluate the efficacy of digital escape rooms as innovative pedagogical tools and to explore their potential in reshaping conventional learning environments.}, language = {en} } @article{AlbrechtRudolph2023, author = {Albrecht, Jens and Rudolph, Eric}, title = {K{\"u}nstliche Intelligenz und Machine Learning}, series = {jugendhilfe}, volume = {61}, journal = {jugendhilfe}, number = {5}, editor = {Dexheimer, Andreas and Rothballer, Marc}, publisher = {Luchterhand}, address = {M{\"u}nchen}, issn = {0022-5940}, pages = {367 - 376}, year = {2023}, subject = {Maschinelles Lernen}, language = {de} } @article{EscobarGrisalesRiosUrregoBaumannetal.2024, author = {Escobar-Grisales, Daniel and R{\´i}os-Urrego, Cristian-David and Baumann, Ilja and Riedhammer, Korbinian and N{\"o}th, Elmar and Bocklet, Tobias and Garcia, Adolfo and Orozco-Arroyave, Juan rafael}, title = {It's Time to Take Action: Acoustic Modeling of Motor Verbs to Detect Parkinson's Disease}, doi = {10.21437/Interspeech.2024-2205}, year = {2024}, abstract = {Pre-trained models generate speech representations that are used in different tasks, including the automatic detection of Parkinson's disease (PD). Although these models can yield high accuracy, their interpretation is still challenging. This paper used a pre-trained Wav2vec 2.0 model to represent speech frames of 25ms length and perform a frame-by-frame discrimination between PD patients and healthy control (HC) subjects. This fine granularity prediction enabled us to identify specific linguistic segments with high discrimination capability. Speech representations of all produced verbs were compared w.r.t. nouns and the first ones yielded higher accuracies. To gaina deeper understanding of this pattern, representations of motor and non-motor verbs were compared and the first ones yielded better results, with accuracies of around 83\% in an independent test set. These findings support well-established neurocognitive models about action-related language highlighted as key drivers of PD. Index Terms: computational paralinguistics, interpretability of pre-trained models, action verbs, Parkinson's disease}, language = {en} } @article{AxenieHalilovMainetal.2025, author = {Axenie, Cristian and Halilov, Ertan and Main, Julian and Weiss, David}, title = {Edge neuro-statistical learning for event-based visual motion detection and tracking in roadside safety systems}, series = {Neuromorphic Computing and Engineering}, journal = {Neuromorphic Computing and Engineering}, publisher = {IOP Publishing}, issn = {2634-4386}, doi = {10.1088/2634-4386/adcbcb}, year = {2025}, abstract = {The Vision Zero Program's purpose is to reduce traffic-related fatalities and serious injuries while promoting equitable, safe, and healthy mobility for all. Ultimately, the challenge is to detect pedestrians during the day and especially at night to implement safety measures. The current study introduces an award-winning low-power solution employing neuromorphic visual sensing and hybrid neuro-statistical processing developed by the Technische Hochschule N{\"u}rnberg team for the TinyML Vision Zero San Jose Competition. The solution proposes a novel neuromorphic edge fusion of spiking neural networks and event-based expectation maximization for the detection and tracking of pedestrians and bicyclists. We provide a deployment-ready evaluation of the detection performance along with robustness, energy footprint, and weatherization while emphasizing the advantages of the neuro-statistical edge solution and its city-level scaling capabilities.}, language = {en} } @article{LippertRudolphPoltermannetal.2024, author = {Lippert, Carolyn and Rudolph, Eric and Poltermann, Aleksandra and Engert, Natalie and Lehmann, Robert and Albrecht, Jens}, title = {Generative KI in der beraterischen Ausbildung}, series = {e-beratungsjournal.net - Zeitschrift f{\"u}r Online-Beratung und computervermittelte Kommunikation}, volume = {20}, journal = {e-beratungsjournal.net - Zeitschrift f{\"u}r Online-Beratung und computervermittelte Kommunikation}, number = {1}, issn = {1816-7632}, doi = {10.48341/tcgc-st69}, pages = {41 - 60}, year = {2024}, abstract = {Der Beitrag beleuchtet das Projekt "Virtuelle*r Klient*in (ViKl)" am Institut f{\"u}r E-Beratung der Technischen Hochschule N{\"u}rnberg Georg Simon Ohm, das sich mit der Nutzung eines KI-basierten Chatbots f{\"u}r die Simulation von Klient*innen in der Lehrveranstaltung "Querschnitt Onlineberatung" im Bachelor-Studiengang Soziale Arbeit befasst. Ziel ist es, praxisnahe Erfahrungen zu vermitteln und die F{\"a}higkeiten der Studierenden im digitalen Beratungsgespr{\"a}ch zu verbessern. Durch den Einsatz verschiedener maschineller Lernarchitekturen werden passende Antworten generiert, die zum Gespr{\"a}chsverlauf und zur psychosozialen Problemstellung passen. Die Nutzung einer umfangreichen Datenbasis aus dem Zertifikatskurs "Onlineberatung" sowie ein zielgruppenad{\"a}quates didaktisches Konzept tragen zur Qualit{\"a}t des*der ViKl bei. Die Diskussion umfasst auch zuk{\"u}nftige Entwicklungen in der Onlineberatung, die didaktische Gestaltung von KI-basierten Lehr-Lern-Formaten, sowie Herausforderungen und Vorteile dieser Ans{\"a}tze.}, subject = {Generative KI}, language = {de} }