Fakultät Informatik
Refine
Year of publication
Document Type
- conference proceeding (article) (194)
- Article (39)
- Part of a Book (28)
- Report (11)
- Bachelor-/Diplom Thesis (3)
- Book (3)
- Other (1)
- Preprint (1)
- Working Paper (1)
Reviewed
- Begutachtet/Reviewed (233)
Keywords
- Large Language Models (5)
- pathological speech (5)
- DFG FIP SPEECHES (4)
- Künstliche Intelligenz (4)
- Automotive Security (3)
- KI (3)
- neuropsychological tests (3)
- speech recognition (3)
- Antifragility (2)
- Artificial Intelligence (2)
Institute
- Fakultät Informatik (281)
- Zentrum für Künstliche Intelligenz (KIZ) (73)
- Institut für e-Beratung (16)
- Fakultät Sozialwissenschaften (15)
- Institut für Angewandte Informatik (4)
- Fakultät Maschinenbau und Versorgungstechnik (2)
- Nuremberg Campus of Technology (2)
- Fakultät Betriebswirtschaft (1)
- Fakultät für Elektrotechnik Feinwerktechnik Informationstechnik (1)
- Institut für Chemie, Material- und Produktentwicklung (1)
Modern vehicle systems are increasingly exposed to cybersecurity threats due to growing connectivity, software complexity, and the integration of external services. While standards such as ISO/SAE 21434 provide guidance for threat analysis and risk assessment, constructing security models remains a predominantly manual activity that is difficult to scale and maintain. This challenge limits the systematic incorporation of cybersecurity concerns into model-based engineering processes and hinders early assessment of security risks. This paper explores the idea of automatically generating security models from structured cyber threat intelligence. We present a prototype approach that transforms publicly available attack knowledge from sources such as MITRE ATT&CK and MITRE EMB3D into instances of the Security Abstraction Model (SAM), a domain-specific security metamodel. Our vision is to establish an attack-driven modeling workflow that continuously integrates evolving threat knowledge into model-based engineering environments. As an initial proof of concept, we implemented a generator and applied it to publicly available attack datasets, resulting in the automatic creation of valid security model instances. These early results indicate the feasibility of extensive security model generation and suggest potential benefits for improving the efficiency of cybersecurity analyses. We discuss open challenges, including semantic enrichment, model integration, and outline future research directions toward continuous, data-driven cybersecurity engineering.
Copy editing of fictional texts requires checks for grammar, usage, style, and consistency. A purely symbolic tool can detect many formal conflicts, but it reaches limits when meaning and a larger context are required. Large language models (LLMs) can handle these questions, but their free-text answers are difficult to integrate into an application. This paper presents the development of EpicStyle from a symbolic system into a hybrid platform. The symbolic pipeline still creates tokens, dictionary entries, grammar structures, and more than 800 rule-based checks. It also selects the context for local LLMs, adds stable identifiers, enriches prompts with story-specific dictionary knowledge, and validates structured answers. We evaluate 29 local model configurations with 346 German grammar test cases. The best task-specific rank is 83.8, while the best observed accuracy is 89.6%. Feedback from normal tool use shows a second challenge: even strong models often generate unnecessary comments. The results support a hybrid design in which deterministic checks and LLM-based judgements have separate and complementary roles.
Large Language Models are increasingly used to generate structured engineering artifacts, yet the instruction artifacts that govern this generation are rarely treated as modeling artifacts in their own right. They typically appear as monolithic prompt blocks, schemas, or informal examples. When tools, metamodels, APIs, or domain vocabularies evolve, stable domain concepts, volatile tool details, validators, and examples drift together in a single prompt contract. This paper puts forward the thesis that prompt-level instruction artifacts should be understood as versioned modeling languages: explicit artifacts that define the LLM-facing structure required to obtain tool-consumable outputs. We realize this idea through Modular Meta-Language-defined Structure Instructions (MMLDSI), a modular architecture that decomposes structured instructions into versioned modules with explicit interfaces, profiles, adapters, validators, examples, and deterministic rule-deck resolution. Changes to tools or domain concepts can then be represented at the affected modules and profiles rather than hidden inside whole-prompt rewrites. The approach is demonstrated in VR scene generation and automotive security modeling. Across a balanced 320-artifact VR subset, required-key validity reaches 88.1%, while a stricter prompt-contract audit accepts 80.9% and exposes remaining tool-readiness gaps. An offline monolith-to-modular pilot maps 12 legacy feature families and resolves 343 case-specific decks without dependency gaps. Our contribution is not another prompting technique, but a model-management perspective on instruction artifacts for evolving LLM toolchains.
LLARS: Enabling Domain Expert & Developer Collaboration for LLM Prompting, Generation and Evaluation
(2026)
We demonstrate LLARS (LLM Assisted Research System), an open-source platform that bridges the gap between domain experts and developers for building LLM-based systems. It integrates three tightly connected modules into an end-to-end pipeline: Collaborative Prompt Engineering for real-time co-authoring with version control and instant LLM testing, Batch Generation for configurable output production across user-selected prompts × models × data with cost control and Hybrid Evaluation where human and LLM evaluators jointly assess outputs through diverse assessment methods, with live agreement metrics and provenance analysis to identify the best model–prompt combination for a given use case. New prompts and models are automatically available for batch generation and completed batches can be turned into evaluation scenarios with a single click. Interviews with six domain experts and three developers in online counselling confirmed that LLARS feels intuitive, saves considerable time by keeping everything in one place and makes interdisciplinary collaboration seamless.
This paper presents OnCoCo 1.0, a new public dataset for fine-grained message classification in online counseling. It is based on a new, integrative system of categories, designed to improve the automated analysis of psychosocial online counseling conversations. Existing category systems, predominantly based on Motivational Interviewing (MI), are limited by their narrow focus and dependence on datasets derived mainly from face-to-face counseling. This limits the detailed examination of textual counseling conversations. In response, we developed a comprehensive new coding scheme that differentiates between 38 types of counselor and 28 types of client utterances, and created a labeled dataset consisting of about 2.800 messages from counseling conversations. We fine-tuned several models on our dataset to demonstrate its applicability. The data and models are publicly available to researchers and practitioners. Thus, our work contributes a new type of fine-grained conversational resource to the language resources community, extending existing datasets for social and mental-health dialogue analysis.
For people with visual impairments, navigating and orienting themselves in public spaces is a daily challenge. Although a wide variety of electronic assistive devices already exist, in the past these provided only fairly abstract guidance. With large language models (LLMs), natural-language descriptions of image content are now possible. While there are already approaches that capture images via smart glasses and forward the image content to external large language models to generate image descriptions, these methods raise concerns regarding data privacy and availability. This paper therefore presents a novel approach in which data is processed exclusively locally on the user’s device. To do this, the user captures a camera image via smart glasses, which is forwarded to a connected smartphone. Inference takes place directly on the smartphone using a reduced, local LLM that has been adapted through LoRA fine-tuning using data relevant to people with visual impairments. The generated image description is then played through a speaker in the temple of the smart glasses. For the task-specific fine-tuning, a comprehensive survey of visually impaired people was conducted, and professional mobility trainers were consulted. The study demonstrated the feasibility of fully local processing using smart glasses in combination with a smartphone. The evaluation showed that fine-tuning the small, local model yields improved image descriptions compared to the non-adapted model, some of which outperform a significantly larger external language model.
Monitoring leftover products provides valuable insights that can be used to optimize future production. This is especially important for German bakeries because freshly baked goods have a very short shelf life. Automating this process can reduce labor costs, improve accuracy, and streamline operations. We propose automating this process using an object detection model to identify baked goods from images. However, the large diversity of German baked goods makes fully supervised training prohibitively expensive and limits
scalability. Although open-vocabulary detectors (e.g., OWLv2, Grounding DINO) offer flexibility, we demonstrate that they are insufficient for our task. While motivated by bakeries, our work addresses the broader challenges of deploying computer vision in industries, where tasks are specialized and annotated datasets are scarce. We compile dataset splits with varying supervision levels, covering 19 classes of baked goods. We propose two training workflows to train an object detection model with limited supervision.
First, we combine OWLv2 and Grounding DINO localizationwith image-level supervision to train the model in a weakly supervised manner. Second, we improve viewpoint robustness by fine-tuning on video frames annotated using Segment Anything 2 as a pseudo-label propagation model. Using these workflows, we train YOLOv11 for our detection task due to its favorable speedaccuracy tradeoff. Relying solely on image-level supervision, the model achieves a mean Average Precision (mAP) of 0.91. Finetuning with pseudo-labels raises model performance by 19.3% under non-ideal deployment conditions. Combining these workflows trains a model that surpasses our fully-supervised baseline model under non-ideal deployment conditions, despite relying only on image-level supervision.
Abstract. Dementia and depression are the most prevalent neuropsychiatric
disorders in geriatric populations, and their overlapping symptoms pose major challenges for differential diagnosis. In this study, we investigate open-weights Large Language Models (LLMs) for predicting dementia and depression severity from speech samples collected during standardized history taking interviews with 154 German-speaking subjects. We introduce an observer-based Global Depression Scale (GDS-D) aligned with the established Global Deterioration Scale (GDS), enabling parallel global staging of affective and cognitive symptoms. We compare three LLMs (Mistral 3.1, DeepHermes, Qwen3) in two settings: (1) zeroshot prediction and (2) LLM-based feature extraction for Support Vector Regression, using human and pause-enriched transcripts. Results show that LLMs effectively predict depression severity in zero-shot settings (best MAE of 0.60), while dementia assessment benefits substantially from structured feature extraction (best MAE of 0.78), reducing errors by up to 35% over zero-shot baselines. Pause-enriched transcripts achieve competitive performance with human transcriptions, demonstrating the viability of fully automatic screening pipelines for differential neuropsychiatric assessment.
Early and accessible detection of Alzheimer’s disease (AD) remains a major challenge, as current diagnostic methods often rely on costly and invasive biomarkers. Speech and language analysis has emerged as a promising
non-invasive and scalable approach to detecting cognitive impairment, but research in this area is hindered by the lack of publicly available datasets, especially for languages other than English. This paper introduces the PARLO
Dementia Corpus (PDC), a new multi-center, clinically validated German resource for AD collected across nine academic memory clinics in Germany. The dataset comprises speech recordings from individuals with AD-related
mild cognitive impairment and mild to moderate dementia, as well as cognitively healthy controls. Speech was elicited using a standardized test battery of eight neuropsychological tasks, including confrontation naming, verbal fluency, word repetition, picture description, story reading, and recall tasks. In addition to audio recordings, the dataset includes manually verified transcriptions and detailed demographic, clinical, and biomarker metadata.
Baseline experiments on ASR benchmarking, automated test evaluation, and LLM-based classification illustrate the feasibility of automatic, speech-based cognitive assessment and highlight the diagnostic value of recall-driven
speech production. The PDC thus establishes the first publicly available German benchmark for multi-modal and cross-lingual research on neurodegenerative diseases.