arXiv subjects

Fabian Wolf

Publications and source records attributed to Fabian Wolf.

At least 19 recordsLinked to original sources

arXiv2026

Quantifying the human visual exposome with vision language models

The visual environment is a fundamental yet unquantified determinant of mental health. While the concept of the environmental exposome is well established, current methods rely on coarse geospatial proxies or biased self reports, failing to capture the first person visual context of daily life. We addressed this gap by coupling ecological momentary assessment with vision language models (VLMs) to quantify the semantic richness of human visual experience. Across 2674 participant generated photographs, VLM derived estimates of greenness robustly predicted momentary affect and chronic stress, consistent with established benchmarks. We then developed a semi autonomous large language model (LLM) based pipeline that mined over seven million scientific publications to extract nearly 1000 environmental features empirically linked to mental health. When applied to real world imagery, up to 33 percent of VLM extracted context ratings significantly correlated with affect and stress. These findings establish a scalable objective paradigm for visual exposomics, enabling high throughput decoding of how the visible world is associated with mental health.

cs.AI↗

arXiv2026

Class Visualizations and Activation Atlases for Enhancing Interpretability in Deep Learning-Based Computational Pathology

The rapid adoption of transformer-based models in computational pathology has enabled prediction of molecular and clinical biomarkers from H&E whole-slide images, yet interpretability has not kept pace with model complexity. While attribution- and generative-based methods are common, feature visualization approaches such as class visualizations (CVs) and activation atlases (AAs) have not been systematically evaluated for these models. We developed a visualization framework and assessed CVs and AAs for a transformer-based foundation model across tissue and multi-organ cancer classification tasks with increasing label granularity. Four pathologists annotated real and generated images to quantify inter-observer agreement, complemented by attribution and similarity metrics. CVs preserved recognizability for morphologically distinct tissues but showed reduced separability for overlapping cancer subclasses. In tissue classification, agreement decreased from Fleiss k = 0.75 (scans) to k = 0.31 (CVs), with similar trends in cancer subclass tasks. AAs revealed layer-dependent organization: coarse tissue-level concepts formed coherent regions, whereas finer subclasses exhibited dispersion and overlap. Agreement was moderate for tissue classification (k = 0.58), high for coarse cancer groupings (k = 0.82), and low at subclass level (k = 0.11). Atlas separability closely tracked expert agreement on real images, indicating that representational ambiguity reflects intrinsic pathological complexity. Attribution-based metrics approximated expert variability in low-complexity settings, whereas perceptual and distributional metrics showed limited alignment. Overall, concept-level feature visualization reveals structured morphological manifolds in transformer-based pathology models and provides a framework for expert-centered interrogation of learned representations across label granularities.

cs.CV↗

arXiv2025

Static Laboratory-Frame Polarization of a Trapped Molecular Ion for CP-Violation Searches

Today's most sensitive experiments for detecting CP-violating permanent electric dipole moments (EDM) rely on molecular spectroscopy. The high sensitivity arises from large internal electric fields that interact with the constituents of the molecule. For molecular ions it has long been assumed that experiments with static polarization from dc electric fields are infeasible, as the ion's charge would either shift it to a field free region or eject it from the trap. This constraint appears to make single ion quantum-logic clocks, among the most precise measurement devices available, incompatible with EDM measurements. Here, we demonstrate that, under typical trapping conditions, heavy molecular ions with small $\Omega/\Lambda$-doubling can be polarized by a static electric field while remaining confined in the Paul trap. This effect arises from a cancellation between the electrostatic force and the trap's ponderomotive force, resulting in an equilibrium position where the ion experiences a dc electric field component. Dynamic decoupling allows to implement co-magnetometry schemes in a single molecule and provides long interrogation times. This will allow the operation of a quantum-logic molecular radio-frequency clock in static electric fields, providing sensitivity to EDMs. Working with only a few ions and non-destructive detection techniques opens the door to the use of highly sensitive, rare, and radioactive molecular species as well as quantum-enhanced metrology schemes to achieve unprecedented levels of accuracy and precision.

physics.atom-ph↗

arXiv2025

Land\'e $g$ factor measurement of $^{48}$Ti$^+$ using simultaneous co-magnetometry and quantum logic spectroscopy

The use of atomic systems as accurate magnetic field probes requires precise characterization of the particle's magnetic properties. Insufficient knowledge of the spatial and temporal characteristics of the external magnetic field often limits the determination of the corresponding atomic parameters. Here, we present a quantum logic scheme mitigating systematic effects caused by temporal magnetic field fluctuations through simultaneous co-magnetometry. This allows measurement of the ground state $g$ factors of a single $^{48}$Ti$^+$ ion with uncertainties at the $10^{-6}$ level. We compare experimentally determined $g$ factors with new theoretical predictions using a combination of configuration interaction (CI) and second-order many-body perturbation theory (MBPT). Theory and experiment agree within the expected level of accuracy. The scheme can be applied to many atomic species, including those that cannot be directly laser cooled.

physics.atom-ph↗

arXiv2025

CM1 -- A Dataset for Evaluating Few-Shot Information Extraction with Large Vision Language Models

The automatic extraction of key-value information from handwritten documents is a key challenge in document analysis. A reliable extraction is a prerequisite for the mass digitization efforts of many archives. Large Vision Language Models (LVLM) are a promising technology to tackle this problem especially in scenarios where little annotated training data is available. In this work, we present a novel dataset specifically designed to evaluate the few-shot capabilities of LVLMs. The CM1 documents are a historic collection of forms with handwritten entries created in Europe to administer the Care and Maintenance program after World War Two. The dataset establishes three benchmarks on extracting name and birthdate information and, furthermore, considers different training set sizes. We provide baseline results for two different LVLMs and compare performances to an established full-page extraction model. While the traditional full-page model achieves highly competitive performances, our experiments show that when only a few training samples are available the considered LVLMs benefit from their size and heavy pretraining and outperform the classical approach.

cs.CV↗

arXiv2024

Quantum logic control of a transition metal ion

Extending quantum control to increasingly complex systems is crucial for both advancing quantum technologies and fundamental physics. In trapped ion systems, quantum logic techniques that combine a well-controlled logic species with a more complex spectroscopy species have proven to be a powerful tool for extending the range of accessible species. Here, we demonstrate that a quantum system as complex as $^{48}$Ti$^+$ with its many metastable states can be controlled employing a combination of intrinsic thermalization due to collisions with background gas and quantum-logic techniques using a far-detuned Raman laser. The preparation of pure quantum states allows coherent manipulation and high resolution measurements of the Zeeman structure in $^{48}$Ti$^+$. The presented techniques are applicable to a wide range of ionic species giving access to a larger variety of systems for fundamental physics and constitute the first step for quantum-controlled spectroscopy of transition metals, relevant, e.g., for the interpretation of astrophysical spectra.

physics.atom-ph↗

arXiv2024

Exploring Architectures for CNN-Based Word Spotting

The goal in word spotting is to retrieve parts of document images which are relevant with respect to a certain user-defined query. The recent past has seen attribute-based Convolutional Neural Networks take over this field of research. As is common for other fields of computer vision, the CNNs used for this task are already considerably deep. The question that arises, however, is: How complex does a CNN have to be for word spotting? Are increasingly deeper models giving increasingly better results or does performance behave asymptotically for these architectures? On the other hand, can similar results be obtained with a much smaller CNN? The goal of this paper is to give an answer to these questions. Therefore, the recently successful TPP-PHOCNet will be compared to a Residual Network, a Densely Connected Convolutional Network and a LeNet architecture empirically. As will be seen in the evaluation, a complex model can be beneficial for word spotting on harder tasks such as the IAM Offline Database but gives no advantage for easier benchmarks such as the George Washington Database.

cs.CV↗

arXiv2024

Deterministic preparation of a dual-species two-ion crystal

The demand for efficient preparation methods for dual-species ion crystals is rapidly expanding across quantum technology and fundamental physics applications with trapped ions. We present a deterministic and efficient technique to produce such crystals, utilizing the segmented structure of a linear Paul trap. By precisely tailoring the trapping potentials, we can split, move, and discard parts of an ion chain. This process is automated in a sequence that converts a larger ion sample into the desired configuration. A critical component of our approach is the accurate identification of crystal constituents. This is achieved by matching the measured positions of fluorescing ions against theoretical expectations for larger crystals, thus facilitating the detection of non-fluorescing ions and enabling accurate ion counting. We demonstrate that our method reliably produces two-ion crystals within minutes. These results represent a significant advance in the production of two-species ion crystals with applications ranging from quantum logic spectroscopy and optical clocks to quantum computing and simulation with trapped ions.

physics.atom-ph↗

arXiv2024

Prospect for precision quantum logic spectroscopy of vibrational overtone transitions in molecular oxygen ions

Precision spectroscopy has been the driving force for progress of our physical understanding and still is a promising tool for the investigation of new physics. Molecules offer transitions which allow tests that are not possible in atomic systems. However, usually precision spectroscopy of molecules is challenging due to the lack of cycling transitions for state preparation and state detection. For molecular ions, this obstacle can be overcome by quantum logic spectroscopy, where dissipation for state preparation and detection is provided by a co-trapped atomic ion exploiting the shared eigenstates of motion. Here, we propose a full quantum logic spectroscopy scheme for molecular oxygen ions and theoretically investigate the feasibility of quantum logic assisted state detection and preparation. Furthermore, we provide coupling rates for a direct single-photon quadrupole excitation of a vibrational overtone transition that can serve as a sensitive transition for tests of a possible variation of the proton-to-electron mass ratio.

physics.atom-ph↗

arXiv2023

A scheme for quantum-logic based transfer of accuracy in polarizability measurement for trapped ions using a moving optical lattice

Optical atomic clocks based on trapped ions suffer from systematic frequency shifts of the clock transition due to interaction with blackbody radiation from the environment. These shifts can be compensated if the blackbody radiation spectrum and the differential dynamic polarizability is known to a sufficient precision. Here, we present a new measurement scheme, based on quantum logic that allows a direct transfer of precision for polarizability measurements from one species to the other. This measurement circumvents the necessity of calibrating laser power below the percent level, which is the limitation for state-of-the-art polarizability measurements in trapped ions. Furthermore, the presented technique allows to reference the polarizability transfer to hydrogen-like ions for which the polarizability can be calculated with high precision.

physics.atom-ph↗

arXiv2022

Self-Training of Handwritten Word Recognition for Synthetic-to-Real Adaptation

Performances of Handwritten Text Recognition (HTR) models are largely determined by the availability of labeled and representative training samples. However, in many application scenarios labeled samples are scarce or costly to obtain. In this work, we propose a self-training approach to train a HTR model solely on synthetic samples and unlabeled data. The proposed training scheme uses an initial model trained on synthetic data to make predictions for the unlabeled target dataset. Starting from this initial model with rather poor performance, we show that a considerable adaptation is possible by training against the predicted pseudo-labels. Moreover, the investigated self-training strategy does not require any manually annotated training samples. We evaluate the proposed method on four widely used benchmark datasets and show its effectiveness on closing the gap to a model trained in a fully-supervised manner.

cs.CV↗

arXiv2022

Recognition-free Question Answering on Handwritten Document Collections

In recent years, considerable progress has been made in the research area of Question Answering (QA) on document images. Current QA approaches from the Document Image Analysis community are mainly focusing on machine-printed documents and perform rather limited on handwriting. This is mainly due to the reduced recognition performance on handwritten documents. To tackle this problem, we propose a recognition-free QA approach, especially designed for handwritten document image collections. We present a robust document retrieval method, as well as two QA models. Our approaches outperform the state-of-the-art recognition-free models on the challenging BenthamQA and HW-SQuAD datasets.

cs.CV↗

arXiv2020

Annotation-free Learning of Deep Representations for Word Spotting using Synthetic Data and Self Labeling

Word spotting is a popular tool for supporting the first exploration of historic, handwritten document collections. Today, the best performing methods rely on machine learning techniques, which require a high amount of annotated training material. As training data is usually not available in the application scenario, annotation-free methods aim at solving the retrieval task without representative training samples. In this work, we present an annotation-free method that still employs machine learning techniques and therefore outperforms other learning-free approaches. The weakly supervised training scheme relies on a lexicon, that does not need to precisely fit the dataset. In combination with a confidence based selection of pseudo-labeled training samples, we achieve state-of-the-art query-by-example performances. Furthermore, our method allows to perform query-by-string, which is usually not the case for other annotation-free methods.

cs.CV↗

arXiv2020

Initialization of quantum simulators by sympathetic cooling

Simulating computationally intractable many-body problems on a quantum simulator holds great potential to deliver insights into physical, chemical, and biological systems. While the implementation of Hamiltonian dynamics within a quantum simulator has already been demonstrated in many experiments, the problem of initialization of quantum simulators to a suitable quantum state has hitherto remained mostly unsolved. Here, we show that already a single dissipatively driven auxiliary particle can efficiently prepare the quantum simulator in a low-energy state of largely arbitrary Hamiltonians. We demonstrate the scalability of our approach and show that it is robust against unwanted sources of decoherence. While our initialization protocol is largely independent of the physical realization of the simulation device, we provide an implementation example for a trapped ion quantum simulator.

quant-ph↗

arXiv2019

Exploring Confidence Measures for Word Spotting in Heterogeneous Datasets

In recent years, convolutional neural networks (CNNs) took over the field of document analysis and they became the predominant model for word spotting. Especially attribute CNNs, which learn the mapping between a word image and an attribute representation, showed exceptional performances. The drawback of this approach is the overconfidence of neural networks when used out of their training distribution. In this paper, we explore different metrics for quantifying the confidence of a CNN in its predictions, specifically on the retrieval problem of word spotting. With these confidence measures, we limit the inability of a retrieval list to reject certain candidates. We investigate four different approaches that are either based on the network's attribute estimations or make use of a surrogate model. Our approach also aims at answering the question for which part of a dataset the retrieval system gives reliable results. We further show that there exists a direct relation between the proposed confidence measures and the quality of an estimated attribute representation.

cs.CV↗

arXiv2018

Motional Fock states for quantum-enhanced amplitude and phase measurements with trapped ions

Non-vanishing fluctuations of the vacuum state are a salient feature of quantum theory. These fluctuations fundamentally limit the precision of quantum sensors. Nowadays, several systems such as optical clocks, gravitational wave detectors, matter-wave interferometers, magnetometers, and optomechanical systems approach measurement sensitivities where the effect of quantum fluctuations sets a fundamental limit, the so-called standard quantum limit (SQL). It has been proposed that the SQL can be overcome by squeezing the vacuum fluctuations. Realizations of this scheme have been demonstrated in various systems. However, protocols based on squeezed vacuum crucially rely on precise control of the relative orientation of the squeezing with respect to the operation imprinting the measured quantity. Lack of control can lead to an amplification of noise and reduces the sensitivity of the device. Here, we experimentally demonstrate a novel quantum metrological paradigm based on phase insensitive Fock states of the motional state of a trapped ion, with applications in frequency metrology and displacement detection. The measurement apparatus is used in two different experimental settings probing non-commuting observables with sensitivities beyond the SQL. In both measurements, classical preparation and detection noise are sufficiently small to preserve the quantum gain in a full metrological protocol.

physics.atom-ph↗

arXiv2015

Non-destructive state detection for quantum logic spectroscopy of molecular ions

Laser spectroscopy of cold and trapped molecular ions is a powerful tool for fundamental physics, including the determination of fundamental constants, the laboratory test for their possible variation, and the search for a possible electric dipole moment of the electron. Optical clocks based on molecular ions sensitive to some of these effects are expected to achieve uncertainties approaching the $10^{-18}$ level. While the complexity of molecular structure facilitates these applications, the absence of cycling transitions poses a challenge for direct laser cooling, quantum state control, and detection. Previously employed state detection techniques based on photo-dissociation or chemical reactions are destructive and therefore inefficient. Here we experimentally demonstrate non-destructive state detection of a single trapped molecular ion through its strong Coulomb coupling to a well-controlled co-trapped atomic ion. An algorithm based on a state-dependent optical dipole force(ODF) changes the internal state of the atom conditioned on the internal state of the molecule. We show that individual states in the molecule can be distinguished by their coupling strength to the ODF and observe black-body radiation-induced quantum jumps between rotational states. Using the detuning dependence of the state detection signal, we implement a variant of quantum logic spectroscopy and improve upon a previous measurement of the $\mathrm{X}^1Σ^+(J=1)\rightarrow\mathrm{A}^1Σ^+(J=0)$ transition in MgH, finding a frequency of 1067.74752(53)THz. We estimate that non-destructive state detection with near 100% efficiency could take less than 10 ms. The technique we demonstrate is applicable to a wide range of molecular ions, enabling further applications in state-controlled quantum chemistry and spectroscopic investigations of molecules serving as probes for interstellar clouds.

physics.atom-ph↗

arXiv2015

Detection of motional ground state population of a trapped ion using delayed pulses

Efficient preparation and detection of the motional state of trapped ions is important in many experiments ranging from quantum computation to precision spectroscopy. We investigate the stimulated Raman adiabatic passage (STIRAP) technique for the manipulation of motional states in a trapped ion system. The presented technique uses a Raman coupling between two hyperfine ground states in $^{25}$Mg$^+$, implemented with delayed pulses, which removes a single phonon independent of the initial motional state. We show that for a thermal state the STIRAP population transfer is more efficient than a stimulated Raman Rabi pulse on a motional sideband. In contrast to previous implementations, a large detuning of more than 200 times the natural linewidth of the transition is used. This approach renders STIRAP suitable for atoms in which resonant laser fields would populate fluorescing excited states and thus impede the STIRAP process. We use the technique to measure the wavefunction overlap of excited motional states with the motional ground state. This is an important application for photon recoil spectroscopy and other force sensing applications that utilize the high sensitivity of the motional state of trapped ions to external fields. Furthermore, a determination of the ground state population enables a simple measurement of the ion's temperature.

quant-ph↗