arXiv subjects

Thomas Fuchs

Publications and source records attributed to Thomas Fuchs.

10 recordsLinked to original sources

arXiv2026

Exploring Post-Training Alignment of Small Language Models for Biomedical Data-to-Text Generation: A Case Study of Medication Leaflet

Translating complex biomedical data into patient-friendly narratives is central to modern biomedical informatics. This study presents a comparative analysis of training small language models (SLMs) in specialized biomedical datato-text generation tasks. We explore widely adopted post-training methods including supervised fine-tuning (SFT), direct preference optimization (DPO), odds ratio preference optimization (ORPO), and group relative policy optimization (GRPO) with Qwen-based SLMs on a medicine package leaflets dataset. To assess cross-dataset generalizability, we also curated drug label data from openFDA. We evaluate models using both standard lexical overlap metrics like ROUGE as well as semantic similarity measures. Across our experiments, the results show that (1) the aligned SLMs outperform proprietary models like GPT-5; (2) ORPO outperforms the SFTbaselines; (3) GRPO yields the most robust cross-dataset performance among the alignment methods tested as well as GPT-5.

cs.CL↗

arXiv2026

The Power of 10: New Rules for the Digital World

As artificial intelligence rapidly advances, society is increasingly captivated by promises of superhuman machines and seamless digital futures. Yet these visions often obscure mounting social, ethical, and psychological concerns tied to pervasive digital technologies - from surveillance to mental health crises. This article argues that a guiding ethos is urgently needed to navigate these transformations. Inspired by the lasting influence of the biblical Ten Commandments, a European interdisciplinary group has proposed "Ten Rules for the Digital World" - a novel ethical framework to help individuals and societies make prudent, human-centered decisions in the age of "supercharged" technology.

cs.CY↗

arXiv2025

Predict Patient Self-reported Race from Skin Histological Images

Artificial Intelligence (AI) has demonstrated success in computational pathology (CPath) for disease detection, biomarker classification, and prognosis prediction. However, its potential to learn unintended demographic biases, particularly those related to social determinants of health, remains understudied. This study investigates whether deep learning models can predict self-reported race from digitized dermatopathology slides and identifies potential morphological shortcuts. Using a multisite dataset with a racially diverse population, we apply an attention-based mechanism to uncover race-associated morphological features. After evaluating three dataset curation strategies to control for confounding factors, the final experiment showed that White and Black demographic groups retained high prediction performance (AUC: 0.799, 0.762), while overall performance dropped to 0.663. Attention analysis revealed the epidermis as a key predictive feature, with significant performance declines when these regions were removed. These findings highlight the need for careful data curation and bias mitigation to ensure equitable AI deployment in pathology. Code available at: https://github.com/sinai-computational-pathology/CPath_SAIF.

cs.CV↗

arXiv2024

Virchow2: Scaling Self-Supervised Mixed Magnification Models in Pathology

Foundation models are rapidly being developed for computational pathology applications. However, it remains an open question which factors are most important for downstream performance with data scale and diversity, model size, and training algorithm all playing a role. In this work, we propose algorithmic modifications, tailored for pathology, and we present the result of scaling both data and model size, surpassing previous studies in both dimensions. We introduce three new models: Virchow2, a 632 million parameter vision transformer, Virchow2G, a 1.9 billion parameter vision transformer, and Virchow2G Mini, a 22 million parameter distillation of Virchow2G, each trained with 3.1 million histopathology whole slide images, with diverse tissues, originating institutions, and stains. We achieve state of the art performance on 12 tile-level tasks, as compared to the top performing competing models. Our results suggest that data diversity and domain-specific methods can outperform models that only scale in the number of parameters, but, on average, performance benefits from the combination of domain-specific methods, data scale, and model scale.

cs.CV↗

arXiv2020

Characterizing the turbulent drag properties of rough surfaces with a Taylor--Couette setup

Wall-roughness induces extra drag in wall-bounded turbulent flows. Mapping any given roughness geometry to its fluid dynamic behaviour has been hampered by the lack of accurate and direct measurements of skin-friction drag. Here the Taylor-Couette (TC) system provides an opportunity as it is a closed system and allows to directly and reliably measure the skin-friction. However, the wall-curvature potentially complicates the connection between the wall friction and the wall roughness characteristics. Here we investigate the effects of a hydrodynamically fully rough surface on highly turbulent, inner cylinder rotating, TC flow. We find that the effects of a hydrodynamically fully rough surface on TC turbulence, where the roughness height k is three orders of magnitude smaller than the Obukhov curvature length Lc (which characterizes the effects of curvature on the turbulent flow, see Berghout et al. arXiv: 2003.03294, 2020), are similar to those effects of a fully rough surface on a flat plate turbulent boundary layer (BL). Hence, the value of the equivalent sand grain height ks, that characterizes the drag properties of a rough surface, is similar to those found for comparable sandpaper surfaces in a flat plate BL. Next, we obtain the dependence of the torque (skin-friction drag) on the Reynolds number for given wall roughness, characterized by ks, and find agreement with the experimental results within 5 percent. Our findings demonstrate that global torque measurements in the TC facility are well suited to reliably deduce wall drag properties for any rough surface.

physics.flu-dyn↗

arXiv2016

Multi-Organ Cancer Classification and Survival Analysis

Accurate and robust cell nuclei classification is the cornerstone for a wider range of tasks in digital and Computational Pathology. However, most machine learning systems require extensive labeling from expert pathologists for each individual problem at hand, with no or limited abilities for knowledge transfer between datasets and organ sites. In this paper we implement and evaluate a variety of deep neural network models and model ensembles for nuclei classification in renal cell cancer (RCC) and prostate cancer (PCa). We propose a convolutional neural network system based on residual learning which significantly improves over the state-of-the-art in cell nuclei classification. Finally, we show that the combination of tissue types during training increases not only classification accuracy but also overall survival analysis.

q-bio.QM↗

arXiv2015

Understanding Neural Networks Through Deep Visualization

Recent years have produced great advances in training large, deep neural networks (DNNs), including notable successes in training convolutional neural networks (convnets) to recognize natural images. However, our understanding of how these models work, especially what computations they perform at intermediate layers, has lagged behind. Progress in the field will be further accelerated by the development of better tools for visualizing and interpreting neural nets. We introduce two such tools here. The first is a tool that visualizes the activations produced on each layer of a trained convnet as it processes an image or video (e.g. a live webcam stream). We have found that looking at live activations that change in response to user input helps build valuable intuitions about how convnets work. The second tool enables visualizing features at each layer of a DNN via regularized optimization in image space. Because previous versions of this idea produced less recognizable images, here we introduce several new regularization methods that combine to produce qualitatively clearer, more interpretable visualizations. Both tools are open source and work on a pre-trained convnet with minimal setup.

cs.CV↗

arXiv2004

Coulomb drag between quantum wires with different electron densities

We study the way back-scattering electron--electron interaction generates Coulomb drag between quantum wires with different densities. At low temperature $T$ the system can undergo a commensurate-- incommensurate transition as the potential difference $|W|$ between the two wires passes a critical value $Δ$, and this transition is reflected in a marked change in the dependence of drag resistivity on $W$ and $T$. At high temperature a density difference between the wires suppresses Coulomb drag induced by back scattering, and we use the Tomonaga--Luttinger model to study this suppression in detail.

cond-mat.mes-hall↗

arXiv2003

Structure of the nucleon in chiral perturbation theory

We discuss a renormalization scheme for relativistic baryon chiral perturbation theory which provides a simple and consistent power counting for renormalized diagrams. The method involves finite subtractions of dimensionally regularized diagrams beyond the standard modified minimal subtraction scheme of chiral perturbation theory to remove contributions violating the power counting. This is achieved by a suitable renormalization of the parameters of the most general effective Lagrangian. As applications we discuss the mass of the nucleon, the $\sigma$ term, and the scalar and electromagnetic form factors.

hep-ph↗

arXiv2003

Electromagnetic form factors of the nucleon in relativistic baryon chiral perturbation theory

The electromagnetic form factors of the nucleon are calculated at order $q^4$ in the relativistic formulation of baryon chiral perturbation theory. In order to obtain a consistent power counting for the renormalized diagrams we make use of the extended on-mass-shell renormalization scheme which we have discussed in a recent paper. We analyze the Dirac and Pauli form factors, $F_1$ and $F_2$, as well as the electric and magnetic Sachs form factors, $G_E$ and $G_M$. Our results are compared with those obtained in the heavy-baryon approach and in the (relativistic) infrared regularization. In addition, the Sachs form factors are compared with experimental data. We confirm previous findings that a one-loop calculation of the electromagnetic form factors does not generate sufficient curvature for the Sachs form factors $G^p_E$, $G^p_M$, and $G^n_M$. Moreover, the electric form factor of the neutron is very sensitive to higher-order contributions.

nucl-th↗