Alexander Löser

Tianyu Han, Lisa C. Adams, Jens-Michalis Papaioannou et al.

As large language models (LLMs) like OpenAI's GPT series continue to make strides, we witness the emergence of artificial intelligence applications in an ever-expanding range of fields. In medicine, these LLMs hold considerable promise for improving medical workflows, diagnostics, patient care, and education. Yet, there is an urgent need for open-source models that can be deployed on-premises to safeguard patient privacy. In our work, we present an innovative dataset consisting of over 160,000 entries, specifically crafted to fine-tune LLMs for effective medical applications. We investigate the impact of fine-tuning these datasets on publicly accessible pre-trained LLMs, and subsequently, we juxtapose the performance of pre-trained-only models against the fine-tuned models concerning the examinations that future medical doctors must pass to achieve certification.

Betty van Aken, Jens-Michalis Papaioannou, Marcel G. Naik et al.

The use of deep neural models for diagnosis prediction from clinical text has shown promising results. However, in clinical practice such models must not only be accurate, but provide doctors with interpretable and helpful results. We introduce ProtoPatient, a novel method based on prototypical networks and label-wise attention with both of these abilities. ProtoPatient makes predictions based on parts of the text that are similar to prototypical patients - providing justifications that doctors understand. We evaluate the model on two publicly available clinical datasets and show that it outperforms existing baselines. Quantitative and qualitative evaluations with medical doctors further demonstrate that the model provides valuable explanations for clinical decision support.

Keno K. Bressem, Jens-Michalis Papaioannou, Paul Grundmann et al.

This paper presents medBERTde, a pre-trained German BERT model specifically designed for the German medical domain. The model has been trained on a large corpus of 4.7 Million German medical documents and has been shown to achieve new state-of-the-art performance on eight different medical benchmarks covering a wide range of disciplines and medical document types. In addition to evaluating the overall performance of the model, this paper also conducts a more in-depth analysis of its capabilities. We investigate the impact of data deduplication on the model's performance, as well as the potential benefits of using more efficient tokenization methods. Our results indicate that domain-specific models such as medBERTde are particularly useful for longer texts, and that deduplication of training data does not necessarily lead to improved performance. Furthermore, we found that efficient tokenization plays only a minor role in improving model performance, and attribute most of the improved performance to the large amount of training data. To encourage further research, the pre-trained model weights and new benchmarks based on radiological data are made publicly available for use by the scientific community.

Jens-Michalis Papaioannou, Paul Grundmann, Betty van Aken et al.

Clinical phenotyping enables the automatic extraction of clinical conditions from patient records, which can be beneficial to doctors and clinics worldwide. However, current state-of-the-art models are mostly applicable to clinical notes written in English. We therefore investigate cross-lingual knowledge transfer strategies to execute this task for clinics that do not use the English language and have a small amount of in-domain data available. We evaluate these strategies for a Greek and a Spanish clinic leveraging clinical notes from different clinical domains such as cardiology, oncology and the ICU. Our results reveal two strategies that outperform the state-of-the-art: Translation-based methods in combination with domain-specific encoders and cross-lingual encoders plus adapters. We find that these strategies perform especially well for classifying rare phenotypes and we advise on which method to prefer in which situation. Our results show that using multilingual data overall improves clinical phenotyping models and can compensate for data sparseness.

Daniil Gurgurov, Tom Röhr, Sebastian von Rohrscheidt et al.

Despite advances in multilingual capabilities, most large language models (LLMs) remain English-centric in their training and, crucially, in their production of reasoning traces. Even when tasked with non-English problems, these models predominantly reason in English, creating a fundamental mismatch for non-English usage scenarios. We address this disparity directly with three contributions. (i) We introduce ReasonXL, the first large-scale parallel corpus of cross-domain reasoning traces spanning five European languages (English, German, French, Italian, and Spanish), with over two million aligned samples per language, each comprising prompts, reasoning traces, and final outputs, enabling direct supervision of language-specific reasoning. (ii) Using ReasonXL, we demonstrate that LLMs can be adapted to reason entirely in a desired target language, using a simple two-stage pipeline of supervised fine-tuning (SFT) followed by reinforcement learning with verifiable rewards (RLVR). The resulting models match or exceed baseline performance, with minimal loss in general knowledge and broadly preserved cross-lingual transfer. (iii) We conduct an extensive representational analysis of the adaptation and find a clear functional division across model depth: early layers contain an activation bottleneck that causally determines language identity, while upper layers concentrate the weight and activation changes driven by adaptation. We further find that RLVR achieves greater behavioral divergence from the base model with smaller parameter updates than SFT, suggesting a more efficient representational rerouting despite much smaller weight updates.

Justus Viga, Penelope Mueck, Alexander Löser et al.

In the shipping industry, fuel consumption and emissions are critical factors due to their significant impact on economic efficiency and environmental sustainability. Accurate prediction of ship fuel consumption is essential for further optimization of maritime operations. However, heterogeneous methodologies and limited high-quality datasets hinder direct comparison of modeling approaches. This paper makes three key contributions: (1) we introduce and release a new dataset (https://huggingface.co/datasets/krohnedigital/FuelCast) comprising operational and environmental data from three ships; (2) we define a standardized benchmark covering tabular regression and time-series regression (3) we investigate the application of in-context learning for ship consumption modeling using the TabPFN foundation model - a first in this domain to our knowledge. Our results demonstrate strong performance across all evaluated models, supporting the feasibility of onboard, data-driven fuel prediction. Models incorporating environmental conditions consistently outperform simple polynomial baselines relying solely on vessel speed. TabPFN slightly outperforms other techniques, highlighting the potential of foundation models with in-context learning capabilities for tabular prediction. Furthermore, including temporal context improves accuracy.

Tom Röhr, Soumyadeep Roy, Fares Al Mohamad et al.

In a doctor-patient dialogue, the primary objective of physicians is to diagnose patients and propose a treatment plan. Medical doctors guide these conversations through targeted questioning to efficiently gather the information required to provide the best possible outcomes for patients. To the best of our knowledge, this is the first work that studies physician intent trajectories in doctor-patient dialogues. We use the `Ambient Clinical Intelligence Benchmark' (Aci-bench) dataset for our study. We collaborate with medical professionals to develop a fine-grained taxonomy of physician intents based on the SOAP framework (Subjective, Objective, Assessment, and Plan). We then conduct a large-scale annotation effort to label over 5000 doctor-patient turns with the help of a large number of medical experts recruited using Prolific, a popular crowd-sourcing platform. This large labeled dataset is an important resource contribution that we use for benchmarking the state-of-the-art generative and encoder models for medical intent classification tasks. Our findings show that our models understand the general structure of medical dialogues with high accuracy, but often fail to identify transitions between SOAP categories. We also report for the first time common trajectories in medical dialogue structures that provide valuable insights for designing `differential diagnosis' systems. Finally, we extensively study the impact of intent filtering for medical dialogue summarization and observe a significant boost in performance. We make the codes and data, including annotation guidelines, publicly available at https://github.com/DATEXIS/medical-intent-classification.

Justus Westerhoff, Golzar Atefi, Mario Koddenbrock et al.

The capacity of a foundation model allows for adaptation to new downstream tasks. Weight imprinting is a universal and efficient method to fulfill this purpose. It has been reinvented several times, but it has not been systematically studied. In this paper, we propose a framework for imprinting, identifying three main components: generation, normalization, and aggregation. This allows us to conduct an in-depth analysis of imprinting and a comparison of the existing work. We reveal the benefits of representing novel data with multiple proxies in the generation step and show the importance of proper normalization. We determine proxies through clustering and propose a novel variant of imprinting that outperforms previous work. We motivate this by the neural collapse phenomenon -- an important connection that we can draw for the first time. Our results show an increase of up to 4\% in challenging scenarios with complex data distributions for new classes. Finally, we publicly release our code at https://github.com/DATEXIS/multi-imprinting/.

Lisa Adams, Felix Busch, Tianyu Han et al.

Background: Recent advancements in large language models (LLMs) offer potential benefits in healthcare, particularly in processing extensive patient records. However, existing benchmarks do not fully assess LLMs' capability in handling real-world, lengthy clinical data. Methods: We present the LongHealth benchmark, comprising 20 detailed fictional patient cases across various diseases, with each case containing 5,090 to 6,754 words. The benchmark challenges LLMs with 400 multiple-choice questions in three categories: information extraction, negation, and sorting, challenging LLMs to extract and interpret information from large clinical documents. Results: We evaluated nine open-source LLMs with a minimum of 16,000 tokens and also included OpenAI's proprietary and cost-efficient GPT-3.5 Turbo for comparison. The highest accuracy was observed for Mixtral-8x7B-Instruct-v0.1, particularly in tasks focused on information retrieval from single and multiple patient documents. However, all models struggled significantly in tasks requiring the identification of missing information, highlighting a critical area for improvement in clinical data interpretation. Conclusion: While LLMs show considerable potential for processing long clinical documents, their current accuracy levels are insufficient for reliable clinical use, especially in scenarios requiring the identification of missing information. The LongHealth benchmark provides a more realistic assessment of LLMs in a healthcare setting and highlights the need for further model refinement for safe and effective clinical application. We make the benchmark and evaluation code publicly available.

Bogdan Kostić, Conor Fallon, Julian Risch et al.

The rapid advancement of Large Language Models (LLMs) has established standardized evaluation benchmarks as the primary instrument for model comparison. Yet, their reliability is increasingly questioned due to sensitivity to shallow variations in input prompts. This paper examines how controlled, truth-conditionally equivalent lexical and syntactic perturbations affect the absolute performance and relative ranking of 23 contemporary LLMs across three benchmarks: MMLU, SQuAD, and AMEGA. We employ two linguistically principled pipelines to generate meaning-preserving variations: one performing synonym substitution for lexical changes, and another using dependency parsing to determine applicable syntactic transformations. Results show that lexical perturbations consistently induce substantial, statistically significant performance degradation across nearly all models and tasks, while syntactic perturbations have more heterogeneous effects, occasionally improving results. Both perturbation types destabilize model leaderboards on complex tasks. Furthermore, model robustness did not consistently scale with model size, revealing strong task dependence. Overall, the findings suggest that LLMs rely more on surface-level lexical patterns than on abstract linguistic competence, underscoring the need for robustness testing as a standard component of LLM evaluation.

Paul Grundmann, Dennis Fast, Jan Frick et al.

With their growing capabilities, generative large language models (LLMs) are being increasingly investigated for complex medical tasks. However, their effectiveness in real-world clinical applications remains underexplored. To address this, we present CliniBench, the first benchmark that enables comparability of well-studied encoder-based classifiers and generative LLMs for discharge diagnosis prediction from admission notes in MIMIC-IV dataset. Our extensive study compares 12 generative LLMs and 3 encoder-based classifiers and demonstrates that encoder-based classifiers consistently outperform generative models in diagnosis prediction. We assess several retrieval augmentation strategies for in-context learning from similar patients and find that they provide notable performance improvements for generative LLMs.

Alexei Figueroa, Justus Westerhoff, Golzar Atefi et al.

Biologically inspired neural networks offer alternative avenues to model data distributions. FlyVec is a recent example that draws inspiration from the fruit fly's olfactory circuit to tackle the task of learning word embeddings. Surprisingly, this model performs competitively even against deep learning approaches specifically designed to encode text, and it does so with the highest degree of computational efficiency. We pose the question of whether this performance can be improved further. For this, we introduce Comply. By incorporating positional information through complex weights, we enable a single-layer neural network to learn sequence representations. Our experiments show that Comply not only supersedes FlyVec but also performs on par with significantly larger state-of-the-art models. We achieve this without additional parameters. Comply yields sparse contextual representations of sentences that can be interpreted explicitly from the neuron weights.

Betty van Aken, Sebastian Herrmann, Alexander Löser

Decision support systems based on clinical notes have the potential to improve patient care by pointing doctors towards overseen risks. Predicting a patient's outcome is an essential part of such systems, for which the use of deep neural networks has shown promising results. However, the patterns learned by these networks are mostly opaque and previous work revealed flaws regarding the reproduction of unintended biases. We thus introduce an extendable testing framework that evaluates the behavior of clinical outcome models regarding changes of the input. The framework helps to understand learned patterns and their influence on model decisions. In this work, we apply it to analyse the change in behavior with regard to the patient characteristics gender, age and ethnicity. Our evaluation of three current clinical NLP models demonstrates the concrete effects of these characteristics on the models' decisions. They show that model behavior varies drastically even when fine-tuned on the same data and that allegedly best-performing models have not always learned the most medically plausible patterns.

Paul Grundmann, Sebastian Arnold, Alexander Löser

Retrieving answer passages from long documents is a complex task requiring semantic understanding of both discourse and document context. We approach this challenge specifically in a clinical scenario, where doctors retrieve cohorts of patients based on diagnoses and other latent medical aspects. We introduce CAPR, a rule-based self-supervision objective for training Transformer language models for domain-specific passage matching. In addition, we contribute a novel retrieval dataset based on clinical notes to simulate this scenario on a large corpus of clinical notes. We apply our objective in four Transformer-based architectures: Contextual Document Vectors, Bi-, Poly- and Cross-encoders. From our extensive evaluation on MIMIC-III and three other healthcare datasets, we report that CAPR outperforms strong baselines in the retrieval of domain-specific passages and effectively generalizes across rule-based and human-labeled passages. This makes the model powerful especially in zero-shot scenarios where only limited training data is available.

Betty van Aken, Jens-Michalis Papaioannou, Manuel Mayrdorfer et al.

Outcome prediction from clinical text can prevent doctors from overlooking possible risks and help hospitals to plan capacities. We simulate patients at admission time, when decision support can be especially valuable, and contribute a novel admission to discharge task with four common outcome prediction targets: Diagnoses at discharge, procedures performed, in-hospital mortality and length-of-stay prediction. The ideal system should infer outcomes based on symptoms, pre-conditions and risk factors of a patient. We evaluate the effectiveness of language models to handle this scenario and propose clinical outcome pre-training to integrate knowledge about patient outcomes from multiple public sources. We further present a simple method to incorporate ICD code hierarchy into the models. We show that our approach improves performance on the outcome tasks against several baselines. A detailed analysis reveals further strengths of the model, including transferability, but also weaknesses such as handling of vital values and inconsistencies in the underlying data.

Betty van Aken, Benjamin Winter, Alexander Löser et al.

Explainability and interpretability are two important concepts, the absence of which can and should impede the application of well-performing neural networks to real-world problems. At the same time, they are difficult to incorporate into the large, black-box models that achieve state-of-the-art results in a multitude of NLP tasks. Bidirectional Encoder Representations from Transformers (BERT) is one such black-box model. It has become a staple architecture to solve many different NLP tasks and has inspired a number of related Transformer models. Understanding how these models draw conclusions is crucial for both their improvement and application. We contribute to this challenge by presenting VisBERT, a tool for visualizing the contextual token representations within BERT for the task of (multi-hop) Question Answering. Instead of analyzing attention weights, we focus on the hidden states resulting from each encoder block within the BERT model. This way we can observe how the semantic representations are transformed throughout the layers of the model. VisBERT enables users to get insights about the model's internal state and to explore its inference steps or potential shortcomings. The tool allows us to identify distinct phases in BERT's transformations that are similar to a traditional NLP pipeline and offer insights during failed predictions.

Sebastian Arnold, Betty van Aken, Paul Grundmann et al.

We present Contextual Discourse Vectors (CDV), a distributed document representation for efficient answer retrieval from long healthcare documents. Our approach is based on structured query tuples of entities and aspects from free text and medical taxonomies. Our model leverages a dual encoder architecture with hierarchical LSTM layers and multi-task training to encode the position of clinical entities and aspects alongside the document discourse. We use our continuous representations to resolve queries with short latency using approximate nearest neighbor search on sentence level. We apply the CDV model for retrieving coherent answer passages from nine English public health resources from the Web, addressing both patients and medical professionals. Because there is no end-to-end training data available for all application scenarios, we train our model with self-supervised data from Wikipedia. We show that our generalized model significantly outperforms several state-of-the-art baselines for healthcare passage ranking and is able to adapt to heterogeneous domains without additional fine-tuning.

Betty van Aken, Benjamin Winter, Alexander Löser et al.

Bidirectional Encoder Representations from Transformers (BERT) reach state-of-the-art results in a variety of Natural Language Processing tasks. However, understanding of their internal functioning is still insufficient and unsatisfactory. In order to better understand BERT and other Transformer-based models, we present a layer-wise analysis of BERT's hidden states. Unlike previous research, which mainly focuses on explaining Transformer models by their attention weights, we argue that hidden states contain equally valuable information. Specifically, our analysis focuses on models fine-tuned on the task of Question Answering (QA) as an example of a complex downstream task. We inspect how QA models transform token vectors in order to find the correct answer. To this end, we apply a set of general and QA-specific probing tasks that reveal the information stored in each representation layer. Our qualitative analysis of hidden state visualizations provides additional insights into BERT's reasoning process. Our results show that the transformations within BERT go through phases that are related to traditional pipeline tasks. The system can therefore implicitly incorporate task-specific information into its token representations. Furthermore, our analysis reveals that fine-tuning has little impact on the models' semantic abilities and that prediction errors can be recognized in the vector representations of even early layers.

Sebastian Arnold, Rudolf Schneider, Philippe Cudré-Mauroux et al.

When searching for information, a human reader first glances over a document, spots relevant sections and then focuses on a few sentences for resolving her intention. However, the high variance of document structure complicates to identify the salient topic of a given section at a glance. To tackle this challenge, we present SECTOR, a model to support machine reading systems by segmenting documents into coherent sections and assigning topic labels to each section. Our deep neural network architecture learns a latent topic embedding over the course of a document. This can be leveraged to classify local topics from plain text and segment a document at topic shifts. In addition, we contribute WikiSection, a publicly available dataset with 242k labeled sections in English and German from two distinct domains: diseases and cities. From our extensive evaluation of 20 architectures, we report a highest score of 71.6% F1 for the segmentation and classification of 30 topics from the English city domain, scored by our SECTOR LSTM model with bloom filter embeddings and bidirectional segmentation. This is a significant improvement of 29.5 points F1 compared to state-of-the-art CNN classifiers with baseline segmentation.

Betty van Aken, Julian Risch, Ralf Krestel et al.

Toxic comment classification has become an active research field with many recently proposed approaches. However, while these approaches address some of the task's challenges others still remain unsolved and directions for further research are needed. To this end, we compare different deep learning and shallow approaches on a new, large comment dataset and propose an ensemble that outperforms all individual models. Further, we validate our findings on a second dataset. The results of the ensemble enable us to perform an extensive error analysis, which reveals open challenges for state-of-the-art methods and directions towards pending future research. These challenges include missing paradigmatic context and inconsistent dataset labels.

Iurii Chernushenko, Felix A. Gers, Alexander Löser et al.

We present a new methodology for high-quality labeling in the fashion domain with crowd workers instead of experts. We focus on the Aspect-Based Sentiment Analysis task. Our methods filter out inaccurate input from crowd workers but we preserve different worker labeling to capture the inherent high variability of the opinions. We demonstrate the quality of labeled data based on Facebook's FastText framework as a baseline.

Torsten Kilias, Alexander Löser, Felix A. Gers et al.

We present a novel architecture, In-Database Entity Linking (IDEL), in which we integrate the analytics-optimized RDBMS MonetDB with neural text mining abilities. Our system design abstracts core tasks of most neural entity linking systems for MonetDB. To the best of our knowledge, this is the first defacto implemented system integrating entity-linking in a database. We leverage the ability of MonetDB to support in-database-analytics with user defined functions (UDFs) implemented in Python. These functions call machine learning libraries for neural text mining, such as TensorFlow. The system achieves zero cost for data shipping and transformation by utilizing MonetDB's ability to embed Python processes in the database kernel and exchange data in NumPy arrays. IDEL represents text and relational data in a joint vector space with neural embeddings and can compensate errors with ambiguous entity representations. For detecting matching entities, we propose a novel similarity function based on joint neural embeddings which are learned via minimizing pairwise contrastive ranking loss. This function utilizes a high dimensional index structures for fast retrieval of matching entities. Our first implementation and experiments using the WebNLG corpus show the effectiveness and the potentials of IDEL.

Rudolf Schneider, Tom Oberhauser, Tobias Klatt et al.

We report results on benchmarking Open Information Extraction (OIE) systems using RelVis, a toolkit for benchmarking Open Information Extraction systems. Our comprehensive benchmark contains three data sets from the news domain and one data set from Wikipedia with overall 4522 labeled sentences and 11243 binary or n-ary OIE relations. In our analysis on these data sets we compared the performance of four popular OIE systems, ClausIE, OpenIE 4.2, Stanford OpenIE and PredPatt. In addition, we evaluated the impact of five common error classes on a subset of 749 n-ary tuples. From our deep analysis we unreveal important research directions for a next generation of OIE systems.

Sebastian Arnold, Felix A. Gers, Torsten Kilias et al.

Named entity recognition often fails in idiosyncratic domains. That causes a problem for depending tasks, such as entity linking and relation extraction. We propose a generic and robust approach for high-recall named entity recognition. Our approach is easy to train and offers strong generalization over diverse domain-specific language, such as news documents (e.g. Reuters) or biomedical text (e.g. Medline). Our approach is based on deep contextual sequence learning and utilizes stacked bidirectional LSTM networks. Our model is trained with only few hundred labeled sentences and does not rely on further external knowledge. We report from our results F1 scores in the range of 84-94% on standard datasets.