Search

Quantum Approximate Optimization Algorithm (QAOA) and Quantum Annealing are prominent approaches for solving combinatorial optimization problems, such as those formulated as Quadratic Unconstrained Binary Optimization (QUBO). These algorithms aim to minimize the objective function $x^T Q x$, where $Q$ is a QUBO matrix. However, the number of two-qubit CNOT gates in QAOA circuits and the complexity of problem embeddings in Quantum Annealing scale linearly with the number of non-zero couplings in $Q$, contributing to significant computational and error-related challenges. To address this, we introduce the concept of \textit{semi-symmetries} in QUBO matrices and propose an algorithm for identifying and factoring these symmetries into ancilla qubits. \textit{Semi-symmetries} frequently arise in optimization problems such as \textit{Maximum Clique}, \textit{Hamilton Cycles}, \textit{Graph Coloring}, and \textit{Graph Isomorphism}. We theoretically demonstrate that the modified QUBO matrix $Q_{\text{mod}}$ retains the same energy spectrum as the original $Q$. Experimental evaluations on the aforementioned problems show that our algorithm reduces the number of couplings and QAOA circuit depth by up to $45\%$. For Quantum Annealing, these reductions also lead to sparser problem embeddings, shorter qubit chains and better performance. This work highlights the utility of exploiting QUBO matrix structure to optimize quantum algorithms, advancing their scalability and practical applicability to real-world combinatorial problems., Comment: arXiv admin note: substantial text overlap with arXiv:2411.08824

Reinforcement Learning (RL) policies are designed to predict actions based on current observations to maximize cumulative future rewards. In real-world applications (i.e., non-simulated environments), sensors are essential for measuring the current state and providing the observations on which RL policies rely to make decisions. A significant challenge in deploying RL policies in real-world scenarios is handling sensor dropouts, which can result from hardware malfunctions, physical damage, or environmental factors like dust on a camera lens. A common strategy to mitigate this issue is the use of backup sensors, though this comes with added costs. This paper explores the optimization of backup sensor configurations to maximize expected returns while keeping costs below a specified threshold, C. Our approach uses a second-order approximation of expected returns and includes penalties for exceeding cost constraints. We then optimize this quadratic program using Tabu Search, a meta-heuristic algorithm. The approach is evaluated across eight OpenAI Gym environments and a custom Unity-based robotic environment (RobotArmGrasping). Empirical results demonstrate that our quadratic program effectively approximates real expected returns, facilitating the identification of optimal sensor configurations., Comment: Accepted at ICAART conference 2025

The performance of large language models (LLMs) is closely linked to their underlying size, leading to ever-growing networks and hence slower inference. Speculative decoding has been proposed as a technique to accelerate autoregressive generation, leveraging a fast draft model to propose candidate tokens, which are then verified in parallel based on their likelihood under the target model. While this approach guarantees to reproduce the target output, it incurs a substantial penalty: many high-quality draft tokens are rejected, even when they represent objectively valid continuations. Indeed, we show that even powerful draft models such as GPT-4o, as well as human text cannot achieve high acceptance rates under the standard verification scheme. This severely limits the speedup potential of current speculative decoding methods, as an early rejection becomes overwhelmingly likely when solely relying on alignment of draft and target. We thus ask the following question: Can we adapt verification to recognize correct, but non-aligned replies? To this end, we draw inspiration from the LLM-as-a-judge framework, which demonstrated that LLMs are able to rate answers in a versatile way. We carefully design a dataset to elicit the same capability in the target model by training a compact module on top of the embeddings to produce ``judgements" of the current continuation. We showcase our strategy on the Llama-3.1 family, where our 8b/405B-Judge achieves a speedup of 9x over Llama-405B, while maintaining its quality on a large range of benchmarks. These benefits remain present even in optimized inference frameworks, where our method reaches up to 141 tokens/s for 8B/70B-Judge and 129 tokens/s for 8B/405B on 2 and 8 H100s respectively.

Short-distance constraints from the operator product expansion in QCD play an important role in the evaluation of the hadronic light-by-light scattering contribution to the anomalous magnetic moment of the muon. While conventional hadronic models involving a finite number of resonances fail to reproduce the correct power laws implied by them, holographic QCD has been shown to naturally incorporate the Melnikov-Vainshtein constraint on the longitudinal amplitude following from the triangle anomaly in the asymmetric limit, where one photon virtuality remains small compared to the others. This is saturated by an infinite tower of axial vector mesons, and their numerical contribution to the muon $g-2$ in AdS/QCD models agrees rather well with a recent dispersive analysis and alternative approaches. However, the longitudinal short-distance constraint where all virtualities are large turns out to be matched only at the level of 81\%. In this Letter we show that tensor mesons, whose contribution to the muon $g-2$ has recently been found to be underestimated, can fill this gap, because in holographic QCD their infinite tower of excited mesons only contributes to the symmetric longitudinal short-distance constraint. Numerically, they give rise to a sizeable positive contribution from the low-energy region below 1.5 GeV, a small one from the mixed region, and a negligible one from the high-energy region, which could explain the remaining gap between the most recent dispersive and lattice results for the complete hadronic light-by-light contribution., Comment: REVTEX, 7 pages, 2 figures, 1 table. arXiv admin note: text overlap with arXiv:2501.09699

In recent years, significant advancements in computational methods have dramatically enhanced the precision in determining the energetic ranking of different phases of molecular crystals. The developments mainly focused on providing accurate dispersion corrected exchange correlation functionals and methods for describing the vibrational entropy contributions to the free energy at finite temperatures. Several molecular crystals phases were recently found to have of multi-minima character. For our investigations we highlight the multi-minima character in the example of the molecular crystal consisting of N-(4-Methylbenzylidene)-4-methylalanine. We explore its potential energy landscape on the full DFT level or with a machine learned potential that was fitted to DFT data. We calculate not only many local minima but also exact barriers along transformation pathways to demonstrate the multi-minima character of our system. Furthermore, we present a framework, based on the quantum superposition method, that includes both configurational and vibrational entropy. As an example, we show for our system that the transition temperature between two of its phases is afflicted by an error of about 200 K if the multi-minima character is not taken into account. This indicates that it is absolutely essential to consider configurational entropy to obtain reliable finite temperature free energy rankings for complex molecular crystals.

Coupled optical cavities, which support normal modes, play a critical role in optical filtering, sensing, slow-light generation, and quantum state manipulation. Recent theoretical work has proposed incorporating nonlinear materials into these systems to enable novel quantum technologies. Here, we report the first experimental demonstration of squeezing generated in a quantum-enhanced coupled-cavity system, achieving a quantum noise reduction of 3.5 dB at a normal-mode splitting frequency of 7.47 MHz. We provide a comprehensive analysis of the system's loss mechanisms and performance limitations, validating theoretical predictions. Our results underscore the promise of coupled-cavity squeezers for advanced quantum applications, including gravitational wave detection and precision sensing.

Pathology Foundation Models (FMs) hold great promise for healthcare. Before they can be used in clinical practice, it is essential to ensure they are robust to variations between medical centers. We measure whether pathology FMs focus on biological features like tissue and cancer type, or on the well known confounding medical center signatures introduced by staining procedure and other differences. We introduce the Robustness Index. This novel robustness metric reflects to what degree biological features dominate confounding features. Ten current publicly available pathology FMs are evaluated. We find that all current pathology foundation models evaluated represent the medical center to a strong degree. Significant differences in the robustness index are observed. Only one model so far has a robustness index greater than one, meaning biological features dominate confounding features, but only slightly. A quantitative approach to measure the influence of medical center differences on FM-based prediction performance is described. We analyze the impact of unrobustness on classification performance of downstream models, and find that cancer-type classification errors are not random, but specifically attributable to same-center confounders: images of other classes from the same medical center. We visualize FM embedding spaces, and find these are more strongly organized by medical centers than by biological factors. As a consequence, the medical center of origin is predicted more accurately than the tissue source and cancer type. The robustness index introduced here is provided with the aim of advancing progress towards clinical adoption of robust and reliable pathology FMs.

We consider gray-box fuzzing of a program instrumented such that information about evaluation of program expressions converting values of numerical types to Boolean, like x <= y, is recorded during each program's execution. Given that information for an executed program path, we formally define the problem for finding input such that program's execution with that input evaluates all those expressions in the same order and with the same Boolean values as in the original execution path, except for the last one, which is evaluated to the opposite value. Then we also provide an algorithm searching for a solution of the problem effectively. The effectiveness of the algorithm is demonstrated empirically via its evaluation on the TestComp 2024 benchmark suite.

Recent hardware advancements in AI Accelerators and GPUs allow to efficiently compute sparse matrix multiplications, especially when 2 out of 4 consecutive weights are set to zero. However, this so-called 2:4 sparsity usually comes at a decreased accuracy of the model. We derive a regularizer that exploits the local correlation of features to find better sparsity masks in trained models. We minimize the regularizer jointly with a local squared loss by deriving the proximal operator for which we show that it has an efficient solution in the 2:4-sparse case. After optimizing the mask, we use maskedgradient updates to further minimize the local squared loss. We illustrate our method on toy problems and apply it to pruning entire large language models up to 70B parameters. On models up to 13B we improve over previous state of the art algorithms, whilst on 70B models we match their performance.

Over recent years, the blockchain ecosystem has grown significantly with the emergence of new Layer-1 (L1) and Layer-2 (L2) networks. These blockchains typically host Decentralized Exchanges (DEXes) for trading assets such as native currencies and stablecoins. While this diversity enriches the ecosystem, it also fragments liquidity, posing challenges for DEXes offering the same assets across multiple blockchains. This fragmentation leads to price discrepancies, creating opportunities like arbitrages for profit-seeking traders, which fall under the broader category of exploitative economic practices known as Maximal Extractable Value (MEV). Although MEV extraction has been extensively studied within single domains (i.e., individual blockchains), cross-chain arbitrages, a form of cross-domain MEV, have received little attention due to their non-atomic nature, complicating both execution and detection. In this paper, we shed light on opaque cross-chain MEV activities by presenting the first systematic study of two non-atomic cross-chain arbitrage strategies: Sequence-Independent Arbitrage (SIA) and Sequence-Dependent Arbitrage (SDA). The former involves independent, opposite-direction trades across chains, while the latter relies on asset bridges. We analyze the effectiveness of these strategies across nine blockchains over a one-year period from September 2023 to August 2024, identifying 260,808 cross-chain arbitrages, 32.37% of which involve bridging solutions. These arbitrages generated a lower-bound profit of 9,496,115.28 USD from a total traded volume of 465,797,487.98 USD. Additionally, we examine the security implications of cross-chain arbitrages, uncovering centralization among arbitrageurs, network congestion caused by failed transactions, and growing private mempool adoption. Finally, we discuss sequencer incentives and propose a risk-optimized arbitrage strategy.

A graph is locally chordal if, locally at each vertex, it is chordal, i.e. composed of cliques glued along a tree. In addition to extending many classic characterizations of chordal graphs, we show that locally chordal graphs are precisely those graphs $G$ which can be decomposed into cliques arranged in the shape of a high-girth graph. The decomposition in the shape of a high-girth graph simultaneously witnesses that $G$ is, locally at each vertex, composed of cliques glued along a tree. We also show that these decompositions can be computed efficiently. Our analysis of locally chordal graphs is inspired by a recent framework to study local vs. global structure in graphs [DJKK22]., Comment: 57 pages, 4 figures

Like any digital certificate, Verifiable Credentials (VCs) require a way to revoke them in case of an error or key compromise. Existing solutions for VC revocation, most prominently Bitstring Status List, are not viable for many use cases since they leak the issuer's behavior, which in turn leaks internal business metrics. For instance, exact staff fluctuation through issuance and revocation of employee IDs. We introduce CRSet, a revocation mechanism that allows an issuer to encode revocation information for years worth of VCs as a Bloom filter cascade. Padding is used to provide deniability for issuer metrics. Issuers periodically publish this filter cascade on a decentralized storage system. Relying Parties (RPs) can download it to perform any number of revocation checks locally. Compared to existing solutions, CRSet protects the metadata of subject, RPs, and issuer equally. At the same time, it is non-interactive, making it work with wallet devices having limited hardware power and drop-in compatible with existing VC exchange protocols and wallet applications. We present a prototype using the Ethereum blockchain as decentralized storage. The recently introduced blob-carrying transactions, enabling cheaper data writes, allow us to write each CRSet directly to the chain. We built software for issuers and RPs that we successfully tested end-to-end with an existing publicly available wallet agents and the OpenID for Verifiable Credentials protocols. Storage and bandwidth costs paid by issuers and RP are higher than for Bitstring Status List, but still manageable at around 1 MB for an issuer issuing hundreds of thousands of VCs annually and covering decades., Comment: Preprint

QCD jets are considered important probes for quark gluon plasma created in collisions of nuclei at high energies. Their parton showers are significantly altered if they develop inside of a deconfined medium. Hadronization of jets is also thought to be affected by the presence of quarks and gluons. We present a systematic study of the effects of a thermal bath of partons on the hadronization of parton showers. We use the JETSCAPE framework to create parton showers both in vacuum and in a brick of quark gluon plasma. The brick setup allows important parameters, like the size of the plasma as well as the collective flow of partons, to be varied systematically. We hadronize the parton showers using Hybrid Hadronization, which permits shower partons to form strings with thermal partons, or to recombine directly with thermal partons as well as with each other. We find a sizeable amount of interaction of shower partons with thermal partons during hadronization, indicating a natural continuation of the interaction of jet and medium during this stage. The observed effects grow with the size of the medium. Collective flow easily transfers from the thermal partons onto the emerging jet hadrons. We also see a significant change in hadron chemistry as expected in the presence of quark recombination processes., Comment: 12 pages, 6 figures

The purpose of this study is to develop an automated and accurate external camera calibration method for multi-camera systems used in 3D surgical scene reconstruction (3D-SSR), eliminating the need for operator intervention or specialized expertise. The method specifically addresses the problem of limited overlapping fields of view caused by significant variations in optical zoom levels and camera locations. We contribute a novel, fast, and fully automatic calibration method based on the projection of multi-scale markers (MSMs) using a ceiling-mounted projector. MSMs consist of 2D patterns projected at varying scales, ensuring accurate extraction of well distributed point correspondences across significantly different viewpoints and zoom levels. Validation is performed using both synthetic and real data captured in a mock-up OR, with comparisons to traditional manual marker-based methods as well as markerless calibration methods. The method achieves accuracy comparable to manual, operator-dependent calibration methods while exhibiting higher robustness under conditions of significant differences in zoom levels. Additionally, we show that state-of-the-art Structure-from-Motion (SfM) pipelines are ineffective in 3D-SSR settings, even when additional texture is projected onto the OR floor. The use of a ceiling-mounted entry-level projector proves to be an effective alternative to operator-dependent, traditional marker-based methods, paving the way for fully automated 3D-SSR.

Under which condition is quantization optimal? We address this question in the context of the additive uniform noise channel under peak amplitude and power constraints. We compute analytically the capacity-achieving input distribution as a function of the noise level, the average power constraint and the exponent of the power constraint. We found that when the cost constraint is tight and the cost function is concave, the capacity-achieving input distribution is discrete, whereas when the cost function is convex, the support of the capacity-achieving input distribution spans the entire interval., Comment: 6 pages, 3 figures

Directed growth, anisotropic cell shapes, and confinement drive self-organization in multicellular systems. We investigate the influence of particle shape on the distribution and dynamics of nematic microdomains in a minimal in-silico model of proliferating, sterically interacting particles, akin to colonies of rod-shaped bacteria. By introducing continuously tuneable tip variations around a common rod shape with spherical caps, we find that subtle changes significantly impact the emergent dynamics, leading to distinct patterns of microdomain formation and stability. Our analysis reveals separate effects of particle shape and aspect ratio, as well as a transition from exponential to scale-free size distributions, which we recapitulate using an effective master equation model. This allows us to relate differences in microdomain size distributions to different physical mechanisms of microdomain breakup. Our results thereby contribute to the characterization of the effective dynamics in growing aggregates at large and intermediate length scales and the microscopic properties that control it. This could be relevant both for biological self-organization and design strategies for future artificial systems., Comment: 7 pages, 6 figures, minor/formatting updates

Advances in computer vision, particularly in optical image-based 3D reconstruction and feature matching, enable applications like marker-less surgical navigation and digitization of surgery. However, their development is hindered by a lack of suitable datasets with 3D ground truth. This work explores an approach to generating realistic and accurate ex vivo datasets tailored for 3D reconstruction and feature matching in open orthopedic surgery. A set of posed images and an accurately registered ground truth surface mesh of the scene are required to develop vision-based 3D reconstruction and matching methods suitable for surgery. We propose a framework consisting of three core steps and compare different methods for each step: 3D scanning, calibration of viewpoints for a set of high-resolution RGB images, and an optical-based method for scene registration. We evaluate each step of this framework on an ex vivo scoliosis surgery using a pig spine, conducted under real operating room conditions. A mean 3D Euclidean error of 0.35 mm is achieved with respect to the 3D ground truth. The proposed method results in submillimeter accurate 3D ground truths and surgical images with a spatial resolution of 0.1 mm. This opens the door to acquiring future surgical datasets for high-precision applications., Comment: 18 pages, 12 figures. Submitted to the 16th International Conference on Information Processing in Computer-Assisted Interventions (IPCAI 2025)

Superconducting nanowire single-photon detectors (SNSPDs) are indispensable in fields such as quantum science and technology, astronomy, and biomedical imaging, where high detection efficiency, low dark count rates and high timing accuracy are required. Recently, helium (He) ion irradiation was shown to be a promising method to enhance SNSPD performance. Here, we study how changes in the underlying superconducting NbTiN film and the SiO2/Si substrate affect device performance. While irradiated and unirradiated NbTiN films show similar crystallinity, we observe He bubble formation below the SiO2/Si interface and an amorphization of the Si substrate. Both reduce the thermal conductance between the superconducting thin film and the substrate from 210 W/m^2/K^4 to 70 W/m^2/K^4 after irradiation with 2000 ions/nm^2. This effect, combined with the lateral straggle of He ions in the substrate, allows the modification of the superconductor-to-substrate thermal conductance of an SNSPD by selectively irradiating the regions around the nanowire. With this approach, we achieved an increased plateau width of saturating intrinsic detection efficiency of 9.8 uA compared to 3.7 uA after full irradiation. Moreover, the critical current remained similar to that of the unirradiated reference device (59 uA versus 60.1 uA), while full irradiation reduced it to 22.4 uA. Our results suggest that the irradiation-induced reduction of the thermal conductance significantly enhances SNSPD sensitivity, offering a novel approach to locally engineer substrate properties for improved detector performance., Comment: 18 pages, 11 figures

The Plug-and-Charge (PnC) process defined by ISO 15118 standardizes automated Electric Vehicle (EV) charging by enabling automatic installation of credentials and use for authentication between EV and Charge Point (CP). However, the current credential installation process is non-uniform, relies on a complex Public Key Infrastructure (PKI), lacks support for fine-grained authorization parameters, and is not very user-friendly. In this paper, we propose a streamlined approach to the initial charging authorization process by leveraging the OAuth Device Authorization Grant and Rich Authorization Requests. The proposed solution reduces technical complexity, simplifies credential installation, introduces flexible authorization constraints (e.g., time- and cost-based), and facilitates payment through OpenID Connect (OIDC). We present a proof-of-concept implementation along with performance evaluations and conduct a symbolic protocol verification using the Tamarin prover. Furthermore, our approach solves the issue of OAuth's cross-device authorization, making it suitable as a formally proven blueprint in contexts beyond EV charging.

The level of tumour-infiltrating lymphocytes (TILs) is a prognostic factor for patients with (triple-negative) breast cancer (BC). Computational TIL assessment (CTA) has the potential to assist pathologists in this labour-intensive task, but current CTA models rely heavily on many detailed annotations. We propose and validate a fundamentally simpler deep learning based CTA that can be trained in only ten minutes on hundredfold fewer pathologist annotations. We collected whole slide images (WSIs) with TILs scores and clinical data of 2,340 patients with BC from six cohorts including three randomised clinical trials. Morphological features were extracted from whole slide images (WSIs) using a pathology foundation model. Our label-efficient Computational stromal TIL assessment model (ECTIL) directly regresses the TILs score from these features. ECTIL trained on only a few hundred samples (ECTIL-TCGA) showed concordance with the pathologist over five heterogeneous external cohorts (r=0.54-0.74, AUROC=0.80-0.94). Training on all slides of five cohorts (ECTIL-combined) improved results on a held-out test set (r=0.69, AUROC=0.85). Multivariable Cox regression analyses indicated that every 10% increase of ECTIL scores was associated with improved overall survival independent of clinicopathological variables (HR 0.86, p<0.01), similar to the pathologist score (HR 0.87, p<0.001). We demonstrate that ECTIL is highly concordant with an expert pathologist and obtains a similar hazard ratio. ECTIL has a fundamentally simpler design than existing methods and can be trained on orders of magnitude fewer annotations. Such a CTA may be used to pre-screen patients for, e.g., immunotherapy clinical trial inclusion, or as a tool to assist clinicians in the diagnostic work-up of patients with BC. Our model is available under an open source licence (https://github.com/nki-ai/ectil)., Comment: Under review. 54 pages including supplementary materials, 2 main tables, 3 main figures, 14 supplementary figures, 4 supplementary tables

We consider the problem of computing the probability of maximality (PoM) of a Gaussian random vector, i.e., the probability for each dimension to be maximal. This is a key challenge in applications ranging from Bayesian optimization to reinforcement learning, where the PoM not only helps with finding an optimal action, but yields a fine-grained analysis of the action domain, crucial in tasks such as drug discovery. Existing techniques are costly, scaling polynomially in computation and memory with the vector size. We introduce LITE, the first approach for estimating Gaussian PoM with almost-linear time and memory complexity. LITE achieves SOTA accuracy on a number of tasks, while being in practice several orders of magnitude faster than the baselines. This also translates to a better performance on downstream tasks such as entropy estimation and optimal control of bandits. Theoretically, we cast LITE as entropy-regularized UCB and connect it to prior PoM estimators., Comment: accepted in AISTATS 2025

We investigate the impact of latent symmetries on the dynamics of photonic systems and their eigenmodes. Residing solely within the eigenspectral domain, latent symmetries are not visible in real space, yet promise intriguing new ways to engineer the functionality of photonic systems. We study the eigenmodes of a 9-site latent-symmetric photonic network and find that an anti-symmetric input state is fundamentally precluded from populating so-called singlet sites. Furthermore, arbitrary extensions of the system at these sites do not break its latent symmetry. Therefore anti-symmetric excitations cannot leave the initial system, which can be leveraged e.g. for the storage of information. This holds true for both single-photon states, or classical light, as well as both distinguishable and indistinguishable two-photon quantum states. Latent symmetries introduce a powerful new set of tools to the design of systems with desired functionality on any nanophotonic platform, paving the way for applications in photonic information processing., Comment: 10 pages, 4 figures

The rise of integer-valued data, partly driven by the Internet of Things (IoT), has increased demand for efficient compression methods to reduce storage and transmission costs. Existing, speed-oriented methods rely on the ``smaller-numbers-less-bits'' principle, assuming unimodal distributions centered around zero. This assumption is often violated in practice, leading to suboptimal compression. We propose QuaRs, a transformation that reshapes arbitrary distributions into unimodal ones centered around zero, improving compatibility with fast integer compression methods. QuaRs remaps data based on quantiles, assigning smaller magnitudes to frequent values. The method is fast, invertible, and has sub-quadratic complexity. QuaRs enhances compression efficiency, even for challenging distributions, while integrating seamlessly with existing techniques.

We investigate the persistence of embedded eigenvalues for a class of magnetic Laplacians on an infinite cylindrical domain. The magnetic potential is assumed to be $C^2$ and asymptotically periodic along the unbounded direction, with an algebraic decay rate towards a periodic background potential. Under the condition that the embedded eigenvalue of the unperturbed operator lies away from the thresholds of the continuous spectrum, we show that the set of nearby potentials for which the embedded eigenvalue persists forms a smooth manifold of finite and even codimension. The proof employs tools from Floquet theory, exponential dichotomies, and Lyapunov--Schmidt reduction. Additionally, we give an example of a potential which satisfies the assumptions of our main theorem., Comment: 18 pages, 4 figures

The transport of quantum states is a crucial aspect of information processing systems, facilitating operations such as quantum key distribution and inter-component communication within quantum computers. Most quantum networks rely on symmetries to achieve an efficient state transfer. A straightforward way to design such networks is to use spatial symmetries, which severely limits the design space. Our work takes a novel approach to designing photonic networks that do not exhibit any conventional spatial symmetries, yet nevertheless support an efficient transfer of quantum states. Paradoxically, while a perfect transfer efficiency is technically unattainable in these networks, a fidelity arbitrarily close to unity is always reached within a finite time of evolution. Key to this approach are so-called latent, or 'hidden', symmetries, which are embodied in the spectral properties of the network. Latent symmetries substantially expand the design space of quantum networks and hold significant potential for applications in quantum cryptography and secure state transfer. We experimentally realize such a nine-site latent-symmetric network and successfully observe state transfer between two sites with a measured fidelity of 75%. Furthermore, by launching a two-photon state, we show that quantum interference is preserved by the network. This demonstrates that the latent symmetries enable efficient quantum state transfer, while offering greater flexibility in designing quantum networks., Comment: 14 pages, 5 figures; Update: inserted arXiv identifier to related reference

Face recognition systems (FRS) exhibit significant accuracy differences based on the user's gender. Since such a gender gap reduces the trustworthiness of FRS, more recent efforts have tried to find the causes. However, these studies make use of manually selected, correlated, and small-sized sets of facial features to support their claims. In this work, we analyse gender bias in face recognition by successfully extending the search domain to decorrelated combinations of 40 non-demographic facial characteristics. First, we propose a toolchain to effectively decorrelate and aggregate facial attributes to enable a less-biased gender analysis on large-scale data. Second, we introduce two new fairness metrics to measure fairness with and without context. Based on these grounds, we thirdly present a novel unsupervised algorithm able to reliably identify attribute combinations that lead to vanishing bias when used as filter predicates for balanced testing datasets. The experiments show that the gender gap vanishes when images of male and female subjects share specific attributes, clearly indicating that the issue is not a question of biology but of the social definition of appearance. These findings could reshape our understanding of fairness in face biometrics and provide insights into FRS, helping to address gender bias issues.

Background: RETFound, a self-supervised, retina-specific foundation model (FM), showed potential in downstream applications. However, its comparative performance with traditional deep learning (DL) models remains incompletely understood. This study aimed to evaluate RETFound against three ImageNet-pretrained supervised DL models (ResNet50, ViT-base, SwinV2) in detecting ocular and systemic diseases. Methods: We fine-tuned/trained RETFound and three DL models on full datasets, 50%, 20%, and fixed sample sizes (400, 200, 100 images, with half comprising disease cases; for each DR severity class, 100 and 50 cases were used. Fine-tuned models were tested internally using the SEED (53,090 images) and APTOS-2019 (3,672 images) datasets and externally validated on population-based (BES, CIEMS, SP2, UKBB) and open-source datasets (ODIR-5k, PAPILA, GAMMA, IDRiD, MESSIDOR-2). Model performance was compared using area under the receiver operating characteristic curve (AUC) and Z-tests with Bonferroni correction (P<0.05/3). Interpretation: Traditional DL models are mostly comparable to RETFound for ocular disease detection with large datasets. However, RETFound is superior in systemic disease detection with smaller datasets. These findings offer valuable insights into the respective merits and limitation of traditional models and FMs.

Learning to change shape is a fundamental strategy of adaptation and evolution of living organisms, from bacteria and cells to tissues and animals. Human-made materials can also exhibit advanced shape morphing capabilities, but lack the ability to learn. Here, we build metamaterials that can learn complex shape-changing responses using a contrastive learning scheme. By being shown examples of the target shape changes, our metamaterials are able to learn those shape changes by progressively updating internal learning degrees of freedom -- the local stiffnesses. Unlike traditional materials that are designed once and for all, our metamaterials have the ability to forget and learn new shape changes in sequence, to learn multiple shape changes that break reciprocity, and to learn multistable shape changes, which in turn allows them to perform reflex gripping actions and locomotion. Our findings establish metamaterials as an exciting platform for physical learning, which in turn opens avenues for the use of physical learning to design adaptive materials and robots., Comment: 12 pages, 5 figures

The experiment performed by Henry Cavendish to measure the density of the earth, is in numerous textbooks described as a measurement of the universal gravitational constant, G, even if we know that this was not true. In this paper, a study on how common this "myth" is based on the checklist developed by Leite on a total of 84 textbooks. The prevalence of the myth in most textbooks throughout the 20th century indicates a focus on the contemporary interests of authors and the physics community in presenting the development of physics. An explanation of the prevalence of the myth and a different approach, Comment: 14 pages

Federated learning (FL) is a decentralized machine learning paradigm, where multiple clients collaboratively train a global model by exchanging only model updates with the central server without sharing the local data of clients. Due to the large volume of model updates required to be transmitted between clients and the central server, most FL systems are associated with high transfer costs (i.e., communication overhead). This issue is more critical for operational applications in remote sensing (RS), especially when large-scale RS data is processed and analyzed through FL systems with restricted communication bandwidth. To address this issue, we introduce an explanation-guided pruning strategy for communication-efficient FL in the context of RS image classification. Our pruning strategy is defined based on the layerwise relevance propagation (LRP) driven explanations to: 1) efficiently and effectively identify the most relevant and informative model parameters (to be exchanged between clients and the central server); and 2) eliminate the non-informative ones to minimize the volume of model updates. The experimental results on the BigEarthNet-S2 dataset demonstrate that our strategy effectively reduces the number of shared model updates, while increasing the generalization ability of the global model. The code of this work will be publicly available at https://git.tu-berlin.de/rsim/FL-LRP, Comment: Submitted to the IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 2025

Many early universe scenarios predict an enhancement of scalar perturbations at scales currently unconstrained by cosmological probes. These perturbations source gravitational waves (GWs) at second order in perturbation theory, leading to a scalar-induced gravitational wave (SIGW) background. The LISA detector, sensitive to mHz GWs, will be able to constrain curvature perturbations in a new window corresponding to scales $k \in [10^{10}, 10^{14}] \,{\rm Mpc}^{-1}$, difficult to probe otherwise. In this work, we forecast the capabilities of LISA to constrain the source of SIGWs using different approaches: i) agnostic, where the spectrum of curvature perturbations is binned in frequency space; ii) template-based, modeling the curvature power spectrum based on motivated classes of models; iii) ab initio, starting from first-principles model of inflation featuring an ultra-slow roll phase. We compare the strengths and weaknesses of each approach. We also discuss the impact on the SIGW spectrum of non-standard thermal histories affecting the kernels of SIGW emission and non-Gaussianity in the statistics of the curvature perturbations. Finally, we propose simple tests to assess whether the signal is compatible with the SIGW hypothesis. The pipeline used is built into the SIGWAY code., Comment: 95 pages, 31 figures

1) Objectives: This technical report presents a synthetic summary and the principal outcomes of the project R-Vessel-X ("Robust vascular network extraction and understanding within hepatic biomedical images") funded by the French Agence Nationale de la Recherche, and developed between 2019 and 2023. 2) Material and methods: We used datasets and tools publicly available such as IRCAD, Bullitt or VascuSynth toobtain real or synthetic angiographic images. The main contributions lie in the field of 3D angiographic image analysis: filtering, segmentation, modeling and simulation, with a specific focus on the liver. 3) Results: We paid a particular attention to open-source software diffusion of the developed methods, by means of 3D Slicer plugins for the liver anatomy segmentation (SlicerRVXLiverSegmentation) and vesselness filtering (Slicer-RVXVesselnessFilters), and an online demo for the generation of synthetic and realistic vessels in 2D and 3D (OpenCCO). 4) Conclusion: The R-Vessel-X project provided extensive research outcomes, covering various topics related to 3D angiographic image analysis, such as filtering, segmentation, modeling and simulation. We also developed open-source and free softwares so that the research communities in biomedical engineering can use these results in their future research., Comment: Innovation and Research in BioMedical engineering, In press

Despite the prominence of tensor mesons in photon-photon collisions, until recently their contribution to the hadronic light-by-light scattering part of the anomalous magnetic moment of the muon has been estimated at the level of only a few $10^{-12}$, with an almost negligible contribution to the error budget of the standard model prediction. A recent reanalysis within the dispersive approach has found that after resolving the issue of kinematic singularities in previous approaches, a larger result is obtained, a few $10^{-11}$, and with opposite sign as in previous results, when a simple quark model for the transition form factors is employed. In this paper, we present the first complete evaluation of tensor meson contributions within a hard-wall model in holographic QCD, which reproduces surprisingly well mass, two-photon width, and the observed singly virtual transition form factors of the dominant $f_2(1270)$, requiring only that the energy-momentum tensor correlator is matched to the leading OPE result of QCD. Due to a second structure function that is absent in the quark model, the result for $a_\mu$ turns out to be positive instead of negative, and also with a magnitude of a few $10^{-11}$. We discuss both pole and non-pole contributions arising from tensor meson exchanges in the holographic light-by-light scattering amplitude, finding that keeping all contributions improves dramatically the convergence of a sum over excited tensor mesons and avoids unnaturally large contributions from the first few excited modes at low energies. Doing so, the contribution from the lightest tensor mode becomes $a_\mu^T=5.9\times 10^{-11}$ ($5.4\times 10^{-11}$ from virtualities less than 1.5 GeV)., Comment: REVTEX, 18 pages, 1 figure, 3 tables

Typically, the von Neumann entropy of a subsystem increases until it plateaus at the thermal value. Under some circumstances, however, the intermediate value can dwarf the final value, even if the subsystem starts in a pure state. A famous example in the context of the black hole information paradox is the entropy of the Hawking radiation, where this behaviour is dubbed the Page curve. More generally, this is the case for excited systems weakly coupled to cold reservoirs. Here we study the entropy dynamics for Lindbladian evolution, i.e. open quantum systems in weak contact with Markovian reservoirs. This allows us to study the non-equilibrium thermodynamics of the subsystem entropy decrease and link it to Landauer's principle: the entropy decrease must be accompanied by a heat flow out of the system. We give an analytic expression of the entanglement dynamics for a decaying excitation in a two-level system and study it under equilibration of a localised oscillator. In both cases the Page time occurs when half the initial energy has left the system., Comment: 5 pages, 2 figures, comments welcome

PAX (antiProtonic Atom X-ray spectroscopy) is a new experiment with the aim to test strong-field quantum electrodynamics (QED) effects by performing high-precision x-ray spectroscopy of antiprotonic atoms. By utilizing advanced microcalorimeter detection techniques and a low-energy antiproton beam provided by the ELENA ring at CERN, gaseous targets will be used for the creation of antiprotonic atoms, and the measurement of transitions between circular Rydberg states will be conducted with up to two orders of magnitude improved accuracy over previous studies using high-purity germanium detectors. Our approach eliminates the longstanding issue of nuclear uncertainties that have hindered prior studies using highly charged ions, thus enabling direct and purely QED-focused measurements. By precisely probing atomic systems with electric fields up to two orders of magnitude above the Schwinger limit, PAX will test vacuum polarization and second-order QED corrections, opening new frontiers in fundamental physics and uncovering potential pathways to physics beyond the Standard Model.

Shape formation is one of the most thoroughly studied problems in most algorithmic models of programmable matter. However, few existing shape formation algorithms utilize similarities between an initial configuration and a desired target shape. In the hybrid model, an active agent with the computational capabilities of a deterministic finite automaton can form shapes by lifting and placing passive tiles on the triangular lattice. We study the shape reconfiguration problem where the agent needs to move all tiles in an input shape to so-called target nodes, which are distinguishable from other nodes by the agent. We present a worst-case optimal $O(mn)$ algorithm for simply connected target shapes and an $O(n^4)$ algorithm for a large class of target shapes that may contain holes, where $m$ is the initial number of unoccupied target nodes and $n$ is the total number of tiles., Comment: 23 pages, 12 figures

The beta Pictoris system, with its two directly imaged planets beta Pic b and beta Pic c and its well characterised debris disk, is a prime target for detailed characterisation of young planetary systems. Here, we present high-resolution and high-contrast LM band spectroscopy with CRIRES+ of the system, primarily for the purpose of atmospheric characterisation of beta Pic b. We developed methods for determining slit geometry and wavelength calibration based on telluric absorption and emission lines, as well as methods for PSF modelling and subtraction, and artificial planet injection, in order to extract and characterise planet spectra at a high S/N and spectral fidelity. Through cross-correlation with model spectra, we detected H2O absorption for planet b in each of the 13 individual observations spanning four different spectral settings. This provides a clear confirmation of previously detected water absorption, and allowed us to derive an exquisite precision on the rotational velocity of beta Pic b, v_rot = 20.36 +/- 0.31 km/s, which is consistent within error bars with previous determinations. We also observed a tentative H2O cross-correlation peak at the expected position and velocity of planet c; the feature is however not at a statistically significant level. Despite a higher sensitivity to SiO than earlier studies, we do not confirm a tentative SiO feature previously reported for planet b. When combining data from different epochs and different observing modes for the strong H2O feature of planet b, we find that the S/N grows considerably faster when sets of different spectral settings are combined, compared to when multiple data sets of the same spectral setting are combined. This implies that maximising spectral coverage is often more important than maximising integration depth when investigating exoplanetary atmospheres using cross-correlation techniques., Comment: 18 pages, 21 figure, accepted for publication in A&A

Glioblastoma, a highly aggressive brain tumor, poses major challenges due to its poor prognosis and high morbidity rates. Partial differential equation-based models offer promising potential to enhance therapeutic outcomes by simulating patient-specific tumor behavior for improved radiotherapy planning. However, model calibration remains a bottleneck due to the high computational demands of optimization methods like Monte Carlo sampling and evolutionary algorithms. To address this, we recently introduced an approach leveraging a neural forward solver with gradient-based optimization to significantly reduce calibration time. This approach requires a highly accurate and fully differentiable forward model. We investigate multiple architectures, including (i) an enhanced TumorSurrogate, (ii) a modified nnU-Net, and (iii) a 3D Vision Transformer (ViT). The optimized TumorSurrogate achieved the best overall results, excelling in both tumor outline matching and voxel-level prediction of tumor cell concentration. It halved the MSE relative to the baseline model and achieved the highest Dice score across all tumor cell concentration thresholds. Our study demonstrates significant enhancement in forward solver performance and outlines important future research directions.

We formalize a rigorous connection between barren plateaus (BP) in variational quantum algorithms and exponential concentration of quantum kernels for machine learning. Our results imply that recently proposed strategies to build BP-free quantum circuits can be utilized to construct useful quantum kernels for machine learning. This is illustrated by a numerical example employing a provably BP-free quantum neural network to construct kernel matrices for classification datasets of increasing dimensionality without exponential concentration.

Recent research indicates that the use of multiple external representations MERs has the potential to support learning, especially in complex scientific areas, such as quantum physics. In particular, the provision of informationally redundant external representations can have advantageous effects on learning outcomes. This is of special relevance for quantum education, where various external representations are available and their effective use is recognised as crucial to student learning. However, research on the effects of informationally redundant external representations in quantum learning is limited. The present study aims to contribute to the development of effective learning materials by investigating the effects of learning with informationally redundant external representations on students' learning of quantum physics. Using a between-subjects design, 113 students were randomly assigned to one of four learning conditions. The control group learnt with a traditional multimedia learning unit on the behaviour of a single photon in a Mach-Zehnder interferometer. The three intervention groups received redundant essential information in the Dirac formalism, the Bloch sphere, or both. The use of eye tracking enabled insight into the learning process depending on the external representations provided. While the results indicate no effect of the study condition on learning outcomes (content knowledge and cognitive load), the analysis of visual behaviour reveals decreased learning efficiency with the addition of the Bloch sphere to the multimedia learning unit. The results are discussed based on current insight in learning with MERs. The study emphasises the need for careful instructional design to balance the associated cognitive load when learning with informationally redundant external representations., Comment: 15 pages, 5 figures, Eva Rexigel and Linda Qerimi contributed equally to this work

The present study examines the effectiveness of applying Artificial Intelligence methods in an automotive production environment to predict unknown lead times in a non-cycle-controlled production area. Data structures are analyzed to identify contextual features and then preprocessed using one-hot encoding. Methods selection focuses on supervised machine learning techniques. In supervised learning methods, regression and classification methods are evaluated. Continuous regression based on target size distribution is not feasible. Classification methods analysis shows that Ensemble Learning and Support Vector Machines are the most suitable. Preliminary study results indicate that gradient boosting algorithms LightGBM, XGBoost, and CatBoost yield the best results. After further testing and extensive hyperparameter optimization, the final method choice is the LightGBM algorithm. Depending on feature availability and prediction interval granularity, relative prediction accuracies of up to 90% can be achieved. Further tests highlight the importance of periodic retraining of AI models to accurately represent complex production processes using the database. The research demonstrates that AI methods can be effectively applied to highly variable production data, adding business value by providing an additional metric for various control tasks while outperforming current non AI-based systems.

Understanding the movement of the left ventricle myocardium (LVmyo) during the cardiac cycle is essential for assessing cardiac function. One way to model this movement is through a series of deformable image registrations (DIRs) of the LVmyo. Traditional deep learning methods for DIRs, such as those based on convolutional neural networks, often require substantial memory and computational resources. In contrast, implicit neural representations (INRs) offer an efficient approach by operating on any number of continuous points. This study extends the use of INRs for DIR to cardiac computed tomography (CT), focusing on LVmyo registration. To enhance the precision of the registration around the LVmyo, we incorporate the signed distance field of the LVmyo with the Hounsfield Unit values from the CT frames. This guides the registration of the LVmyo, while keeping the tissue information from the CT frames. Our framework demonstrates high registration accuracy and provides a robust method for temporal registration that facilitates further analysis of LVmyo motion., Comment: 9 pages, 5 figures, STACOM 2024

Precise diagnostic on the electron beam parameters is a very valuable tool and essential in the operation of synchrotron light sources. One possible option is to employ the emitted synchrotron radiation for non-destructive measurements. A tool, which has been used in many ring based synchrotron light sources is Time-Correlated Single-Photon Counting (TCSPC). It allows to measure the arrival time distribution of the emitted photons and by that reveals the filling pattern, i.e., the charge distribution onto the electron bunches stored in the storage ring. At MAX IV, two TCSPC setups were installed and the analysis was developed further to also allow for the measurement of the longitudinal profiles of the individual bunches. The analysis is available as a Tango device in the accelerator control system and continuously provides, for example, the bunch length of each bunch as well as the bunch profiles and phases. This improved the diagnostic capabilities significantly, for example, in the presence of Landau cavities, which are becoming increasingly more common in new fourth-generation synchrotron light sources., Comment: 20 pages, 15 figures

The radiative transfer equation models various physical processes ranging from plasma simulations to radiation therapy. In practice, these phenomena are often subject to uncertainties. Modeling and propagating these uncertainties requires accurate and efficient solvers for the radiative transfer equations. Due to the equation's high-dimensional phase space, fine-grid solutions of the radiative transfer equation are computationally expensive and memory-intensive. In recent years, dynamical low-rank approximation has become a popular method for solving kinetic equations due to the development of computationally inexpensive, memory-efficient and robust algorithms like the augmented basis update \& Galerkin integrator. In this work, we propose a low-rank Monte Carlo estimator and combine it with a control variate strategy based on multi-fidelity low-rank approximations for variance reduction. We investigate the error analytically and numerically and find that a joint approach to balance rank and grid size is necessary. Numerical experiments further show that the efficiency of estimators can be improved using dynamical low-rank approximation, especially in the context of control variates., Comment: 16 pages, 2 figures, 1 table, Submitted to Proceedings of MCQMC 2024

Recent advances in digital pathology have demonstrated the effectiveness of foundation models across diverse applications. In this report, we present Atlas, a novel vision foundation model based on the RudolfV approach. Our model was trained on a dataset comprising 1.2 million histopathology whole slide images, collected from two medical institutions: Mayo Clinic and Charit\'e - Universt\"atsmedizin Berlin. Comprehensive evaluations show that Atlas achieves state-of-the-art performance across twenty-one public benchmark datasets, even though it is neither the largest model by parameter count nor by training dataset size.

This paper presents a novel optimization approach for allocating grid operation costs in Peer-to-Peer (P2P) electricity markets using Quantum Computing (QC). We develop a Quadratic Unconstrained Binary Optimization (QUBO) model that matches logical power flows between producer-consumer pairs with the physical power flow to distribute grid usage costs fairly. The model is evaluated on IEEE test cases with up to 57 nodes, comparing Quantum Annealing (QA), hybrid quantum-classical algorithms, and classical optimization approaches. Our results show that while the model effectively allocates grid operation costs, QA performs poorly in comparison despite extensive hyperparameter optimization. The classical branch-and-cut method outperforms all solvers, including classical heuristics, and shows the most advantageous scaling behavior. The findings may suggest that binary least-squares optimization problems may not be suitable candidates for near-term quantum utility.

We present crest, a tool for automatically proving (non-)confluence and termination of logically constrained rewrite systems. We compare crest to other tools for logically constrained rewriting. Extensive experiments demonstrate the promise of crest., Comment: Accepted at the 31st International Conference on Tools and Algorithms for the Construction and Analysis of Systems (TACAS) 2025

We present a novel passivity enforcement (passivation) method, called KLAP, for linear time-invariant systems based on the Kalman-Yakubovich-Popov (KYP) lemma and the closely related Lur'e equations. The passivation problem in our framework corresponds to finding a perturbation to a given non-passive system that renders the system passive while minimizing the $\mathcal{H}_2$ or frequency-weighted $\mathcal{H}_2$ distance between the original non-passive and the resulting passive system. We show that this problem can be formulated as an unconstrained optimization problem whose objective function can be differentiated efficiently even in large-scale settings. We show that any minimizer of the unconstrained problem yields the same passive system. Furthermore, we prove that, in the absence of a feedthrough term, every local minimizer is also a global minimizer. For cases involving a non-trivial feedthrough term, we analyze global minimizers in relation to the extremal solutions of the Lur'e equations, which can serve as tools for identifying local minima. To solve the resulting numerical optimization problem efficiently, we propose an initialization strategy based on modifying the feedthrough term and a restart strategy when it is likely that the optimization has converged to a local minimum. Numerical examples illustrate the effectiveness of the proposed method.

The solid acids CsH$_2$PO$_4$ and Cs$_7$(H$_4$PO$_4$)(H$_2$PO$_4$)$_8$ pose significant challenges for the simulation of proton transport phenomena. In this work, we use the recently developed machine-learned force field (MLFF) MACE to model the proton dynamics on nanosecond time scales for these systems and compare its performance with long-term ab initio molecular dynamics (AIMD) simulations. The MACE-MP-0 foundation model shows remarkable performance for all observables derived from molecular dynamics (MD) simulations, but minor quantitative discrepancies remain compared to the AIMD reference data. However, we show that minimal fine-tuning -- fitting to as little as 1 ps of AIMD data -- leads to full quantitative agreement between the radial distribution functions of MACE force field and AIMD simulations. In addition, we show that traditional long-term AIMD simulations fail to capture the correct qualitative trends in diffusion coefficients and activation energies for these solid acids due to the limited accessible time scale. In contrast, accurate and convergent diffusion coefficients can be reliably obtained through multi-nanosecond long MD simulations using machine-learned force fields. The obtained qualitative and quantitative behavior of the converged diffusion coefficients and activation energies now matches the experimental trends for both solid acids, in contrast to previous AIMD simulations that yielded a qualitatively wrong picture.

Large Language Models (LLMs) created new opportunities for generating personas, which are expected to streamline and accelerate the human-centered design process. Yet, AI-generated personas may not accurately represent actual user experiences, as they can miss contextual and emotional insights critical to understanding real users' needs and behaviors. This paper examines the differences in how users perceive personas created by LLMs compared to those crafted by humans regarding their credibility for design. We gathered ten human-crafted personas developed by HCI experts according to relevant attributes established in related work. Then, we systematically generated ten personas and compared them with human-crafted ones in a survey. The results showed that participants differentiated between human-created and AI-generated personas, with the latter being perceived as more informative and consistent. However, participants noted that the AI-generated personas tended to follow stereotypes, highlighting the need for a greater emphasis on diversity when utilizing LLMs for persona creation.