Search

State-of-the-art quantum simulators permit local temporal control of interactions and midcircuit readout. These capabilities open the way towards the exploration of intriguing nonequilibrium phenomena. We illustrate this with a kinetically constrained many-body quantum system that has a natural implementation on Rydberg quantum simulators. The evolution proceeds in discrete time and is generated by repeatedly entangling the system with an auxiliary environment that is monitored and reset after each time-step. Despite featuring an uncorrelated infinite-temperature average stationary state, the dynamics displays coexistence of fast and slow space-time regions in stochastic realizations of the system state. The time-record of measurement outcomes on the environment serves as natural probe for such dynamical heterogeneity, which we characterize using tools from large deviation theory. Our work establishes the large deviation framework for discrete-time open quantum many-body systems as a means to characterize complex dynamics and collective phenomena in quantum processors and simulators., Comment: 8+6 pages, 3+2 figures, comments welcome

Coupling is an essential mechanism that drives complexity in natural systems, transforming single, non-interacting elements into intricate networks with rich physical properties. Here, we demonstrate a chip-scale coupled laser system that exhibits complex optical states impossible to achieve in an uncoupled system. We show that a pair of coupled semiconductor ring lasers spontaneously forms a frequency comb consisting of the hybridized modes of its coupled cavity, exhibiting a large number of phase-locked tones that anticross with one another. Experimental coherent waveform reconstruction reveals that the hybridized frequency comb manifests itself as pairs of bright and dark picosecond-long solitons circulating simultaneously. The dark and bright solitons exit the coupled cavity at the same time, leading to breathing bright solitons temporally overlapped with their dark soliton counterparts - a state inaccessible for a single, free-running laser. Our results demonstrate that the rules that govern allowable states of light can be broken by simply coupling elements together, paving the way for the design of more complex networks of coupled on-chip lasers., Comment: 13 pages, 3 figures

The problem of quantum state estimation is crucial in the development of quantum technologies. In particular, the use of symmetric quantum states is useful in many relevant applications. In this work, we analyze the task of reconstructing the density matrices of symmetric quantum states generated by a quantum processor. For this purpose, we take advantage of an estimation technique that results to be equivalent to the quantum Maximum Entropy (MaxEnt) estimation, and which was recently adapted to quantum states with arbitrary symmetries. The smart use of prior knowledge of the quantum state symmetries allows for a reduction in both, the number of measurements that need to be made on the system, and the size of the computational problem to store and process the data, resulting in a better overall performance of the estimator as well. After performing numerical simulations, we implement some examples of symmetric states in IonQ quantum processors, and estimate them using the proposed technique. The results are in a good agreement with numerical simulations, showing that the proposed method is a good estimator that allows to save both, experimental and computational resources.

Visual retrieval systems face significant challenges when updating models with improved representations due to misalignment between the old and new representations. The costly and resource-intensive backfilling process involves recalculating feature vectors for images in the gallery set whenever a new model is introduced. To address this, prior research has explored backward-compatible training methods that enable direct comparisons between new and old representations without backfilling. Despite these advancements, achieving a balance between backward compatibility and the performance of independently trained models remains an open problem. In this paper, we address it by expanding the representation space with additional dimensions and learning an orthogonal transformation to achieve compatibility with old models and, at the same time, integrate new information. This transformation preserves the original feature space's geometry, ensuring that our model aligns with previous versions while also learning new data. Our Orthogonal Compatible Aligned (OCA) approach eliminates the need for re-indexing during model updates and ensures that features can be compared directly across different model updates without additional mapping functions. Experimental results on CIFAR-100 and ImageNet-1k demonstrate that our method not only maintains compatibility with previous models but also achieves state-of-the-art accuracy, outperforming several existing methods., Comment: Accepted at BEW2024 Workshop at ECCV2024

Recently, foundation models based on Vision Transformers (ViTs) have become widely available. However, their fine-tuning process is highly resource-intensive, and it hinders their adoption in several edge or low-energy applications. To this end, in this paper we introduce an efficient fine-tuning method for ViTs called $\textbf{ALaST}$ ($\textit{Adaptive Layer Selection Fine-Tuning for Vision Transformers}$) to speed up the fine-tuning process while reducing computational cost, memory load, and training time. Our approach is based on the observation that not all layers are equally critical during fine-tuning, and their importance varies depending on the current mini-batch. Therefore, at each fine-tuning step, we adaptively estimate the importance of all layers and we assign what we call ``compute budgets'' accordingly. Layers that were allocated lower budgets are either trained with a reduced number of input tokens or kept frozen. Freezing a layer reduces the computational cost and memory usage by preventing updates to its weights, while discarding tokens removes redundant data, speeding up processing and reducing memory requirements. We show that this adaptive compute allocation enables a nearly-optimal schedule for distributing computational resources across layers, resulting in substantial reductions in training time (up to 1.5x), FLOPs (up to 2x), and memory load (up to 2x) compared to traditional full fine-tuning approaches. Additionally, it can be successfully combined with other parameter-efficient fine-tuning methods, such as LoRA.

The Ricordi archive, a prestigious collection of significant musical manuscripts from renowned opera composers such as Donizetti, Verdi and Puccini, has been digitized. This process has allowed us to automatically extract samples that represent various musical elements depicted on the manuscripts, including notes, staves, clefs, erasures, and composer's annotations, among others. To distinguish between digitization noise and actual music elements, a subset of these images was meticulously grouped and labeled by multiple individuals into several classes. After assessing the consistency of the annotations, we trained multiple neural network-based classifiers to differentiate between the identified music elements. The primary objective of this study was to evaluate the reliability of these classifiers, with the ultimate goal of using them for the automatic categorization of the remaining unannotated data set. The dataset, complemented by manual annotations, models, and source code used in these experiments are publicly accessible for replication purposes., Comment: Accepted at AudioMostly 2024

Individuals engaging on social media often tend to establish online communities where interactions predominantly occur among like-minded peers. While considerable efforts have been devoted to studying and delineating these communities, there has been limited attention directed towards individuals who diverge from these patterns. In this study, we examine the community structure of re-post networks within the context of a polarized political environment at two different times. We specifically identify individuals who consistently switch between opposing communities and analyze the key features that distinguish them. Our investigation focuses on two crucial aspects of these users: the topological properties of their interactions and the political bias in the content of their posts. Our analysis is based on a dataset comprising 2 million tweets related to US President Donald Trump, coupled with data from over 100 000 individual user accounts spanning the 2020 US presidential election year. Our findings indicate that individuals who switch communities exhibit disparities compared to those who remain within the same communities, both in terms of the topological aspects of their interaction patterns (pagerank, degree, betweenness centrality.) and in the sentiment bias of their content towards Donald Trump.

Supervised machine learning (ML) is used extensively in biology and deserves closer scrutiny. The DOME recommendations aim to enhance the validation and reproducibility of ML research by establishing standards for key aspects such as data handling and processing, optimization, evaluation, and model interpretability. The recommendations help to ensure that key details are reported transparently by providing a structured set of questions. Here, we introduce the DOME Registry (URL: registry.dome-ml.org), a database that allows scientists to manage and access comprehensive DOME-related information on published ML studies. The registry uses external resources like ORCID, APICURON and the Data Stewardship Wizard to streamline the annotation process and ensure comprehensive documentation. By assigning unique identifiers and DOME scores to publications, the registry fosters a standardized evaluation of ML methods. Future plans include continuing to grow the registry through community curation, improving the DOME score definition and encouraging publishers to adopt DOME standards, promoting transparency and reproducibility of ML in the life sciences.

The growing concerns regarding energy consumption and privacy have prompted the development of AI solutions deployable on the edge, circumventing the substantial CO2 emissions associated with cloud servers and mitigating risks related to sharing sensitive data. But deploying Convolutional Neural Networks (CNNs) on non-off-the-shelf edge devices remains a complex and labor-intensive task. In this paper, we present and end-to-end workflow for deployment of CNNs on Field Programmable Gate Arrays (FPGAs) using the Gemmini accelerator, which we modified for efficient implementation on FPGAs. We describe how we leverage the use of open source software on each optimization step of the deployment process, the customizations we added to them and its impact on the final system's performance. We were able to achieve real-time performance by deploying a YOLOv7 model on a Xilinx ZCU102 FPGA with an energy efficiency of 36.5 GOP/s/W. Our FPGA-based solution demonstrates superior power efficiency compared with other embedded hardware devices, and even outperforms other FPGA reference implementations. Finally, we present how this kind of solution can be integrated into a wider system, by testing our proposed platform in a traffic monitoring scenario., Comment: 8 pages, 9 figures, accepted at the 27th Euromicro Conference Series on Digital System Design (DSD) 2024

Slate recommendation is a technique commonly used on streaming platforms and e-commerce sites to present multiple items together. A significant challenge with slate recommendation is managing the complex combinatorial choice space. Traditional methods often simplify this problem by assuming users engage with only one item at a time. However, this simplification does not reflect the reality, as users often interact with multiple items simultaneously. In this paper, we address the general slate recommendation problem, which accounts for simultaneous engagement with multiple items. We propose a generative approach using Diffusion Models, leveraging their ability to learn structures in high-dimensional data. Our model generates high-quality slates that maximize user satisfaction by overcoming the challenges of the combinatorial choice space. Furthermore, our approach enhances the diversity of recommendations. Extensive offline evaluations on applications such as music playlist generation and e-commerce bundle recommendations show that our model outperforms state-of-the-art baselines in both relevance and diversity., Comment: 9 pages, 5 figures, 3 tables

Advances in large language models have raised concerns about their potential use in generating compelling election disinformation at scale. This study presents a two-part investigation into the capabilities of LLMs to automate stages of an election disinformation operation. First, we introduce DisElect, a novel evaluation dataset designed to measure LLM compliance with instructions to generate content for an election disinformation operation in localised UK context, containing 2,200 malicious prompts and 50 benign prompts. Using DisElect, we test 13 LLMs and find that most models broadly comply with these requests; we also find that the few models which refuse malicious prompts also refuse benign election-related prompts, and are more likely to refuse to generate content from a right-wing perspective. Secondly, we conduct a series of experiments (N=2,340) to assess the "humanness" of LLMs: the extent to which disinformation operation content generated by an LLM is able to pass as human-written. Our experiments suggest that almost all LLMs tested released since 2022 produce election disinformation operation content indiscernible by human evaluators over 50% of the time. Notably, we observe that multiple models achieve above-human levels of humanness. Taken together, these findings suggest that current LLMs can be used to generate high-quality content for election disinformation operations, even in hyperlocalised scenarios, at far lower costs than traditional methods, and offer researchers and policymakers an empirical benchmark for the measurement and evaluation of these capabilities in current and future models.

We introduce Imagen 3, a latent diffusion model that generates high quality images from text prompts. We describe our quality and responsibility evaluations. Imagen 3 is preferred over other state-of-the-art (SOTA) models at the time of evaluation. In addition, we discuss issues around safety and representation, as well as methods we used to minimize the potential harm of our models.

Recent surge in Large Language Model (LLM) availability has opened exciting avenues for research. However, efficiently interacting with these models presents a significant hurdle since LLMs often reside on proprietary or self-hosted API endpoints, each requiring custom code for interaction. Conducting comparative studies between different models can therefore be time-consuming and necessitate significant engineering effort, hindering research efficiency and reproducibility. To address these challenges, we present prompto, an open source Python library which facilitates asynchronous querying of LLM endpoints enabling researchers to interact with multiple LLMs concurrently, while maximising efficiency and utilising individual rate limits. Our library empowers researchers and developers to interact with LLMs more effectively and enabling faster experimentation and evaluation. prompto is released with an introductory video (https://youtu.be/-eZAmlV4ypk) under MIT License and is available via GitHub (https://github.com/alan-turing-institute/prompto).

The field of single-cell biology is growing rapidly and is generating large amounts of data from a variety of species, disease conditions, tissues, and organs. Coordinated efforts such as CZI CELLxGENE, HuBMAP, Broad Institute Single Cell Portal, and DISCO, allow researchers to access large volumes of curated datasets. Although the majority of the data is from scRNAseq experiments, a wide range of other modalities are represented as well. These resources have created an opportunity to build and expand the computational biology ecosystem to develop tools necessary for data reuse, and for extracting novel biological insights. Here, we highlight achievements made so far, areas where further development is needed, and specific challenges that need to be overcome.

Tables are crucial containers of information, but understanding their meaning may be challenging. Over the years, there has been a surge in interest in data-driven approaches based on deep learning that have increasingly been combined with heuristic-based ones. In the last period, the advent of \acf{llms} has led to a new category of approaches for table annotation. However, these approaches have not been consistently evaluated on a common ground, making evaluation and comparison difficult. This work proposes an extensive evaluation of four STI SOTA approaches: Alligator (formerly s-elbat), Dagobah, TURL, and TableLlama; the first two belong to the family of heuristic-based algorithms, while the others are respectively encoder-only and decoder-only Large Language Models (LLMs). We also include in the evaluation both GPT-4o and GPT-4o-mini, since they excel in various public benchmarks. The primary objective is to measure the ability of these approaches to solve the entity disambiguation task with respect to both the performance achieved on a common-ground evaluation setting and the computational and cost requirements involved, with the ultimate aim of charting new research paths in the field., Comment: 13 pages, 6 figures; fixed avg. accuracy-over-price plot for GPT families, fixed typos in table referencing, added evaluation and inference subsubsection

The ability to anticipate others' goals and intentions is at the basis of human-human social interaction. Such ability, largely based on non-verbal communication, is also a key to having natural and pleasant interactions with artificial agents, like robots. In this work, we discuss a preliminary experiment on the use of head pose as a visual cue to understand and anticipate action goals, particularly reaching and transporting movements. By reasoning on the spatio-temporal connections between the head, hands and objects in the scene, we will show that short-range anticipation is possible, laying the foundations for future applications to human-robot interaction., Comment: Accepted at the workshop on advancing Group Understanding and robots' adaptive behavior (GROUND), held at the Robotics Science and Systems (RSS) Conference, 2024

Some gas-rich "ultra-diffuse" galaxies (UDGs), which are extreme examples of low surface brightness (LSB) dwarf galaxies, have been reported to lack dark matter and to be offset from the baryonic Tully-Fisher relation (BTFR). If confirmed, these UDGs would represent a serious challenge for both LCDM galaxy-formation models and Milgromian dynamics. Here I demonstrate that these conclusions are very dubious due to underestimated uncertainties on inclinations and/or distances. First, I show that UDGs are offset from the BTFR in the same way as usual face-on LSB dwarfs due to systematic biases at low inclinations. Next, I analyze the two UDGs with the best available rotation-curve data. The first UDG (AGC 242019) is ideally inclined for kinematic studies; MOND can fit the observed rotation curve with a distance of 12.5 +/- 0.6 Mpc, which is consistent with Virgocentric flow models. The second UDG (AGC 114905) is close to face-on, so not ideal for kinematic studies; MOND can fit the observed rotation curve with a distance of 68 +/- 13 Mpc and inclination of 15 +/- 2 degrees, which are consistent with existing data. In particular, I show that the disk inclination is more uncertain than previously estimated due to significant asymmetries (lopsidedness) in the stellar distribution. In conclusion, there is no strong evidence that gas-rich UDGs and gas-rich LSB dwarfs are distinct galaxy populations with different dynamical properties; instead, UDGs seem to be a subset of LSB dwarf galaxies biased toward face-on systems., Comment: 5 pages, 4 figures, accepted for publication in Astronomy & Astrophysics Letters

We compute the electron self-energy in Quantum Electrodynamics to three loops in terms of iterated integrals over kernels of elliptic type. We make use of the differential equations method, augmented by an $\epsilon$-factorized basis, which allows us to gain full control over the differential forms appearing in the iterated integrals to all orders in the dimensional regulator. We obtain compact analytic expressions, for which we provide generalized series expansion representations that allow us to evaluate the result numerically for all values of the electron momentum squared. As a by product, we also obtain $\epsilon$-resummed results for the self-energy in the on-shell limit $p^2 = m^2$, which we use to recompute the known three-loop renormalization constants in the on-shell scheme., Comment: 41 pages, 4 tables, 7 figures

Performance testing aims at uncovering efficiency issues of software systems. In order to be both effective and practical, the design of a performance test must achieve a reasonable trade-off between result quality and testing time. This becomes particularly challenging in Java context, where the software undergoes a warm-up phase of execution, due to just-in-time compilation. During this phase, performance measurements are subject to severe fluctuations, which may adversely affect quality of performance test results. However, these approaches often provide suboptimal estimates of the warm-up phase, resulting in either insufficient or excessive warm-up iterations, which may degrade result quality or increase testing time. There is still a lack of consensus on how to properly address this problem. Here, we propose and study an AI-based framework to dynamically halt warm-up iterations at runtime. Specifically, our framework leverages recent advances in AI for Time Series Classification (TSC) to predict the end of the warm-up phase during test execution. We conduct experiments by training three different TSC models on half a million of measurement segments obtained from JMH microbenchmark executions. We find that our framework significantly improves the accuracy of the warm-up estimates provided by state-of-practice and state-of-the-art methods. This higher estimation accuracy results in a net improvement in either result quality or testing time for up to +35.3% of the microbenchmarks. Our study highlights that integrating AI to dynamically estimate the end of the warm-up phase can enhance the cost-effectiveness of Java performance testing., Comment: Accepted for publication in The 39th IEEE/ACM International Conference on Automated Software Engineering (ASE '24)

The safety of Large Language Models (LLMs) remains a critical concern due to a lack of adequate benchmarks for systematically evaluating their ability to resist generating harmful content. Previous efforts towards automated red teaming involve static or templated sets of illicit requests and adversarial prompts which have limited utility given jailbreak attacks' evolving and composable nature. We propose a novel dynamic benchmark of composable jailbreak attacks to move beyond static datasets and taxonomies of attacks and harms. Our approach consists of three components collectively called h4rm3l: (1) a domain-specific language that formally expresses jailbreak attacks as compositions of parameterized prompt transformation primitives, (2) bandit-based few-shot program synthesis algorithms that generate novel attacks optimized to penetrate the safety filters of a target black box LLM, and (3) open-source automated red-teaming software employing the previous two components. We use h4rm3l to generate a dataset of 2656 successful novel jailbreak attacks targeting 6 state-of-the-art (SOTA) open-source and proprietary LLMs. Several of our synthesized attacks are more effective than previously reported ones, with Attack Success Rates exceeding 90% on SOTA closed language models such as claude-3-haiku and GPT4-o. By generating datasets of jailbreak attacks in a unified formal representation, h4rm3l enables reproducible benchmarking and automated red-teaming, contributes to understanding LLM safety limitations, and supports the development of robust defenses in an increasingly LLM-integrated world. Warning: This paper and related research artifacts contain offensive and potentially disturbing prompts and model-generated content.

Porous media are ubiquitous in the fields of energy storage and conversion, catalysis, biomechanics, hydrogeology, and other chemical engineering processes. These media possess high surface-to-volume ratios and their complex channels can restrict and direct the flow. This makes them key components of multiple equipment despite the challenges in selecting design parameters for specific applications. Pore-resolved CFD reveals the effects of their structure at the microscopic scale, but is currently limited by high computing costs and the performance of mesh generation algorithms. This work introduces a RBF-based representation of solids in a massively parallel immersed-boundary framework, enabling both the usage of non-conformal grids and dynamic mesh adaptation. We verify it using the method of manufactured solutions. We validate it using pressure drop measurements through porous silicone monoliths digitized by X-ray computed microtomography for Reynolds numbers up to 30, using grids of 200 M cells distributed over 8 k cores. The reliable model is then used to highlight that pore network structure is the main factor describing pressure evolution and that preferential channels are present at this scale of the porous media. This work opens the door to design and optimize processes by linking microscopic flow to macroscopic properties through the usage of physics-based digital twins of complex porous media., Comment: 26 pages, 18 figures

The emergence of federated learning (FL) presents a promising approach to leverage decentralized data while preserving privacy. Furthermore, the combination of FL and anomaly detection is particularly compelling because it allows for detecting rare and critical anomalies (usually also rare in locally gathered data) in sensitive data from multiple sources, such as cybersecurity and healthcare. However, benchmarking the performance of anomaly detection methods in FL environments remains an underexplored area. This paper introduces FedAD-Bench, a unified benchmark for evaluating unsupervised anomaly detection algorithms within the context of FL. We systematically analyze and compare the performance of recent deep learning anomaly detection models under federated settings, which were typically assessed solely in centralized settings. FedAD-Bench encompasses diverse datasets and metrics to provide a holistic evaluation. Through extensive experiments, we identify key challenges such as model aggregation inefficiencies and metric unreliability. We present insights into FL's regularization effects, revealing scenarios in which it outperforms centralized approaches due to its inherent ability to mitigate overfitting. Our work aims to establish a standardized benchmark to guide future research and development in federated anomaly detection, promoting reproducibility and fair comparison across studies., Comment: 8 pages, 1 figure

Sequential Recommender Systems (SRSs) have emerged as a highly efficient approach to recommendation systems. By leveraging sequential data, SRSs can identify temporal patterns in user behaviour, significantly improving recommendation accuracy and relevance.Ensuring the reproducibility of these models is paramount for advancing research and facilitating comparisons between them. Existing works exhibit shortcomings in reproducibility and replicability of results, leading to inconsistent statements across papers. Our work fills these gaps by standardising data pre-processing and model implementations, providing a comprehensive code resource, including a framework for developing SRSs and establishing a foundation for consistent and reproducible experimentation. We conduct extensive experiments on several benchmark datasets, comparing various SRSs implemented in our resource. We challenge prevailing performance benchmarks, offering new insights into the SR domain. For instance, SASRec does not consistently outperform GRU4Rec. On the contrary, when the number of model parameters becomes substantial, SASRec starts to clearly dominate all the other SRSs. This discrepancy underscores the significant impact that experimental configuration has on the outcomes and the importance of setting it up to ensure precise and comprehensive results. Failure to do so can lead to significantly flawed conclusions, highlighting the need for rigorous experimental design and analysis in SRS research. Our code is available at https://github.com/antoniopurificato/recsys_repro_conf., Comment: 8 pages, 5 figures

This paper investigates the impact of using first names in Large Language Models (LLMs) and Vision Language Models (VLMs), particularly when prompted with ethical decision-making tasks. We propose an approach that appends first names to ethically annotated text scenarios to reveal demographic biases in model outputs. Our study involves a curated list of more than 300 names representing diverse genders and ethnic backgrounds, tested across thousands of moral scenarios. Following the auditing methodologies from social sciences we propose a detailed analysis involving popular LLMs/VLMs to contribute to the field of responsible AI by emphasizing the importance of recognizing and mitigating biases in these systems. Furthermore, we introduce a novel benchmark, the Pratical Scenarios Benchmark (PSB), designed to assess the presence of biases involving gender or demographic prejudices in everyday decision-making scenarios as well as practical scenarios where an LLM might be used to make sensible decisions (e.g., granting mortgages or insurances). This benchmark allows for a comprehensive comparison of model behaviors across different demographic categories, highlighting the risks and biases that may arise in practical applications of LLMs and VLMs., Comment: Accepted at ECCV workshop FAILED

In this paper we describe the tangent vectors of the stable and unstable manifold of a class of Anosov diffeomorphisms on the torus $\mathbb{T}^2$ using the method of formal series and derivative trees. We start with linear automorphism that is hyperbolic and whose eigenvectors are orthogonal. Then we study the perturbation of such maps by trigonometric polynomial. It is known that there exist a (continuous) map $H$ which acts as a change of coordinate between the perturbed and unperturbed system, but such a map is in general, not differentiable. By "re-scaling" the parametrization $H$, we will be able to obtain the explicit formula for the tangent vectors of these maps.

Motivated by the authors' work on permuto-associahedra, which can be considered as a symmetrization of the associahedron using the symmetric group, we introduce and study the $\mathfrak{G}$-symmetrization of an arbitrary polytope $P$ for any reflection group $\mathfrak{G}$. We show that the combinatorics, and moreover, the normal fan of such a symmetrization can be recovered from its refined fundamental fan, a decorated poset describing how the normal fan of $P$ subdivides the fundamental chamber associated to the reflection group $\mathfrak{G}$. One important application of our results is providing a way to approach the realization problem of a $\mathfrak{G}$-symmetric poset F, that is, the problem of constructing a polytope whose face poset is F. Instead of working with the original poset F, we look at its dual poset T (which is $\mathfrak{G}$-symmetric as well) and focus on a generating subposet Z of T, and reduce the problem to realizing Z as a refined fundamental fan., Comment: 33 pages, 12 figures, comments are welcome

This paper develops sharp testable implications for Tobit and IV-Tobit models' identifying assumptions: linear index specification, (joint) normality of latent errors, and treatment (instrument) exogeneity and relevance. The new sharp testable equalities can detect all possible observable violations of the identifying conditions. We propose a testing procedure for the model's validity using existing inference methods for intersection bounds. Simulation results suggests proper size for large samples and that the test is powerful to detect large violation of the exogeneity assumption and violations in the error structure. Finally, we review and propose new alternative paths to partially identify the parameters of interest under less restrictive assumptions.

Switchability of materials properties by applying controlled stimuli such as voltage pulses is an emerging field of study with applicability in adaptive and programmable devices like neuromorphic transistors or non-emissive smart displays. One of the most exciting approaches to modulate materials performance is mobile ion/vacancy insertion for inducing changes in relevant electrical, optical, or magnetic properties, among others. Unveiling the interplay between changes in the concentration of mobile defects (like oxygen vacancies) and functional properties in relevant materials represents a step forward for underpinning the emerging oxide iontronics discipline. In this work, electrochemical oxide-ion pumping cells were fabricated for an analog control of the oxygen stoichiometry in thin films of mixed ionic-electronic conductor La0.5Sr0.5FeO3-{\delta}. We demonstrate over more than 4 orders of magnitude electronic conductivity control within the same crystallographic phase through the precise and continuous voltage control of the oxygen stoichiometry. We show that behind the modification of the transport properties of the material lays a paramagnetic-to-antiferromagnetic transition. We exploit such magnetoelectric coupling to show control over the exchange interaction between La0.5Sr0.5FeO3-{\delta} and a ferromagnetic Co layer deposited on top.

We report the first result from a dark photon dark matter search in the mass range from ${78.62}$ to $83.95~\mathrm{\mu eV}/c^2$ with a dielectric haloscope prototype for MADMAX (Magnetized Disc and Mirror Axion eXperiment). Putative dark photons would convert to observable photons within a stack consisting of three sapphire disks and a mirror. The emitted power of this system is received by an antenna and successively digitized using a low-noise receiver. No dark photon signal has been observed. Assuming unpolarized dark photon dark matter with a local density of $\rho_{\chi}=0.3~\mathrm{GeV/cm^3}$ we exclude a dark photon to photon mixing parameter $\chi > 3.0 \times 10^{-12}$ over the full mass range and $\chi > 1.2 \times 10^{-13}$ at a mass of $80.57~\mathrm{\mu eV}/c^2$ with a 95\% confidence level. This is the first physics result from a MADMAX prototype and exceeds previous constraints on $\chi$ in this mass range by up to almost three orders of magnitude., Comment: v1

In this study, we aim to determine if generalized sounds and music can share a common emotional space, improving predictions of emotion in terms of arousal and valence. We propose the use of multiple datasets as a multi-domain learning technique. Our approach involves creating a common space encompassing features that characterize both generalized sounds and music, as they can evoke emotions in a similar manner. To achieve this, we utilized two publicly available datasets, namely IADS-E and PMEmo, following a standardized experimental protocol. We employed a wide variety of features that capture diverse aspects of the audio structure including key parameters of spectrum, energy, and voicing. Subsequently, we performed joint learning on the common feature space, leveraging heterogeneous model architectures. Interestingly, this synergistic scheme outperforms the state-of-the-art in both sound and music emotion prediction. The code enabling full replication of the presented experimental pipeline is available at https://github.com/LIMUNIMI/MusicSoundEmotions., Comment: Accepted at Audio Mostly 2024, Milan

Aerospace manufacturing companies, such as Thales Alenia Space, design, develop, integrate, verify, and validate products characterized by high complexity and low volume. They carefully document all phases for each product but analyses across products are challenging due to the heterogeneity and unstructured nature of the data in documents. In this paper, we propose a hybrid methodology that leverages Knowledge Graphs (KGs) in conjunction with Large Language Models (LLMs) to extract and validate data contained in these documents. We consider a case study focused on test data related to electronic boards for satellites. To do so, we extend the Semantic Sensor Network ontology. We store the metadata of the reports in a KG, while the actual test results are stored in parquet accessible via a Virtual Knowledge Graph. The validation process is managed using an LLM-based approach. We also conduct a benchmarking study to evaluate the performance of state-of-the-art LLMs in executing this task. Finally, we analyze the costs and benefits of automating preexisting processes of manual data extraction and validation for subsequent cross-report analyses., Comment: Paper Accepted at ISWC 2024 In-Use Track

The graph coloring problem is an optimization problem involving the assignment of one of q colors to each vertex of a graph such that no two adjacent vertices share the same color. This problem is NP-hard and arises in various practical applications. In this work, we present a novel algorithm that leverages graph neural networks to tackle the problem efficiently, particularly for large graphs. We propose a physics-inspired approach that leverages tools used in statistical mechanics to improve the training and performance of the algorithm. The scaling of our method is evaluated for different connectivities and graph sizes. Finally, we demonstrate the effectiveness of our method on a dataset of Erdos-Renyi graphs, showing its applicability also in hard-to-solve connectivity regions where traditional methods struggle., Comment: 16 pages, 9 figures

We consider the problem of transfer of energy to high frequencies in a quasilinear Schr\"odinger equation with sublinear dispersion, on the one dimensional torus. We exhibit initial data undergoing finite but arbitrary large Sobolev norm explosion: their initial norm is arbitrary small in Sobolev spaces of high regularity, but at a later time becomes arbitrary large. We develop a novel mechanism producing instability, which is based on extracting, via paradifferential normal forms, an effective equation driving the dynamics whose leading term is a non-trivial transport operator with non-constant coefficients. We prove that such operator is responsible for energy cascades via a positive commutator estimate inspired by Mourre's commutator theory.

We propose a two step strategy for estimating one-dimensional dynamical parameters of a quantum Markov chain, which involves quantum post-processing the output using a coherent quantum absorber and a "pattern counting'' estimator computed as a simple additive functional of the outcomes trajectory produced by sequential, identical measurements on the output units. We provide strong theoretical and numerical evidence that the estimator achieves the quantum Cram\'{e}-Rao bound in the limit of large output size. Our estimation method is underpinned by an asymptotic theory of translationally invariant modes (TIMs) built as averages of shifted tensor products of output operators, labelled by binary patterns. For large times, the TIMs form a bosonic algebra and the output state approaches a joint coherent state of the TIMs whose amplitude depends linearly on the mismatch between system and absorber parameters. Moreover, in the asymptotic regime the TIMs capture the full quantum Fisher information of the output state. While directly probing the TIMs' quadratures seems impractical, we show that the standard sequential measurement is an effective joint measurement of all the TIMs number operators; indeed, we show that counts of different binary patterns extracted from the measurement trajectory have the expected joint Poisson distribution. Together with the displaced-null methodology of J. Phys. A: Math. Theor. 57 245304 2024 this provides a computationally efficient estimator which only depends on the total number of patterns. This opens the way for similar estimation strategies in continuous-time dynamics, expanding the results of Phys. Rev. X 13, 031012 2023., Comment: 34 pages, 4 figures. Comments and suggestions are more than welcome!

A new Newtonian limit of general relativity is established for stationary axisymmetric gravitationally bound differentially rotating matter distributions with internal pressure. The self-consistent coupling of quasilocal gravitational energy and angular momentum leads to a modified Poisson equation. The coupled equations of motion of the effective fluid elements are also modified, with quasilocal angular momentum and frame-dragging leading to novel dynamics. The solutions of the full system reproduce the phenonomenology of collisionless dark matter for disc galaxies, offering an explanation for their observed rotation curves. Halos of abundant cold dark matter particles are not required., Comment: 6 pages, 2 figures

We define a motivic Greenlees spectral sequence by characterising an associated $t$-structure. We then examine a motivic version of topological Hochschild homology for the motivic cohomology spectrum modulo a prime number $p$. Finally, we use the motivic Greenlees spectral sequence to determine the homotopy ring of a related spectrum, given that the base field is algebraically closed with a characteristic that is coprime to $p$., Comment: 49 pages, 6 figures

In recent years, deep learning methods have shown impressive results for camera-based remote physiological signal estimation, clearly surpassing traditional methods. However, the performance and generalization ability of Deep Neural Networks heavily depends on rich training data truly representing different factors of variation encountered in real applications. Unfortunately, many current remote photoplethysmography (rPPG) datasets lack diversity, particularly in darker skin tones, leading to biased performance of existing rPPG approaches. To mitigate this bias, we introduce PhysFlow, a novel method for augmenting skin diversity in remote heart rate estimation using conditional normalizing flows. PhysFlow adopts end-to-end training optimization, enabling simultaneous training of supervised rPPG approaches on both original and generated data. Additionally, we condition our model using CIELAB color space skin features directly extracted from the facial videos without the need for skin-tone labels. We validate PhysFlow on publicly available datasets, UCLA-rPPG and MMPD, demonstrating reduced heart rate error, particularly in dark skin tones. Furthermore, we demonstrate its versatility and adaptability across different data-driven rPPG methods.

Radio Frequency (RF) wireless power transfer is a promising technology that has the potential to constantly power small Internet of Things (IoT) devices, enabling even battery-less systems and reducing their maintenance requirements. However, to achieve this ambitious goal, carefully designed RF energy harvesting (EH) systems are needed to minimize the conversion losses and the conversion efficiency of the limited power. For intelligent internet of things sensors and devices, which often have non-constant power requirements, an additional power management stage with energy storage is needed to temporarily provide a higher power output than the power being harvested. This paper proposes an RF wireless power energy conversion system for miniaturized IoT composed of an impedance matching network, a rectifier, and power management with energy storage. The proposed sub-system has been experimentally validated and achieved an overall power conversion efficiency (PCE) of over 30 % for an input power of -10 dBm and a peak efficiency of 57 % at 3 dBm.

Potato yield is an important metric for farmers to further optimize their cultivation practices. Potato yield can be estimated on a harvester using an RGB-D camera that can estimate the three-dimensional (3D) volume of individual potato tubers. A challenge, however, is that the 3D shape derived from RGB-D images is only partially completed, underestimating the actual volume. To address this issue, we developed a 3D shape completion network, called CoRe++, which can complete the 3D shape from RGB-D images. CoRe++ is a deep learning network that consists of a convolutional encoder and a decoder. The encoder compresses RGB-D images into latent vectors that are used by the decoder to complete the 3D shape using the deep signed distance field network (DeepSDF). To evaluate our CoRe++ network, we collected partial and complete 3D point clouds of 339 potato tubers on an operational harvester in Japan. On the 1425 RGB-D images in the test set (representing 51 unique potato tubers), our network achieved a completion accuracy of 2.8 mm on average. For volumetric estimation, the root mean squared error (RMSE) was 22.6 ml, and this was better than the RMSE of the linear regression (31.1 ml) and the base model (36.9 ml). We found that the RMSE can be further reduced to 18.2 ml when performing the 3D shape completion in the center of the RGB-D image. With an average 3D shape completion time of 10 milliseconds per tuber, we can conclude that CoRe++ is both fast and accurate enough to be implemented on an operational harvester for high-throughput potato yield estimation. Our method can also be applied to other tuber, fruit and vegetable crops, thereby enabling versatile, accurate and real-time yield monitoring in precision agriculture. Our code, network weights and dataset are publicly available at https://github.com/UTokyo-FieldPhenomics-Lab/corepp.git., Comment: 20 pages, 11 figures, 6 tables

In arXiv:1712.00555, H. Heine shows that given a symmetric monoidal $\infty$-category $\mathcal{V}$ and a weakly $\mathcal{V}$-enriched monad $T$ over an $\infty$-category $\mathcal{C}$, then there is an induced action of $\mathcal{V}$ on $LMod_T(\mathcal{C})$. Moreover, properties like tensoring or enrichment can be transferred from the action on $\mathcal{C}$ to that on $LMod_T(\mathcal{C})$. We see that the action of an internal operad $O \in Alg(sSeq(\mathcal{C}))$ can be interpreted as the action of a monad $T_O$, such that $Alg_O(\mathcal{C})\cong LMod_{T_O}(\mathcal{C})$. We can then prove that, under a presentability assumption, if the category $\mathcal{C}$ admits cotensors with respect to the action of $\mathcal{V}$, then so does $Alg_O(\mathcal{C})\cong LMod_{T_O}(\mathcal{C})$. This is used to show that the coproduct-idempotent algebras are fixed by the induced tensoring action. We apply this to the stable motivic homotopy category and prove that the tensor of any motivic sphere with rational motivic cohomology is equivalent to the latter., Comment: 32 pages, 1 figure

Binary neutron star mergers are among the most energetic events in our Universe, with magnetic fields significantly impacting their dynamics, particularly after the merger. While numerical-relativity simulations that correctly describe the physics are essential to model their rich phenomenology, the inclusion of magnetic fields is crucial for realistic simulations. For this reason, we have extended the BAM code to enable general relativistic magneto-hydrodynamic (GRMHD) simulations employing a hyperbolic `divergence cleaning' scheme. We present a large set of standard GRMHD tests and compare the BAM code to other GRMHD codes, SPRITZ, GRaM-X, and SACRA$_{\rm KK22}$, which employ different schemes for the evolution of the magnetic fields. Overall, we find that the BAM code shows a good performance in simple special-relativistic tests. In addition, we find good agreement and consistent results when comparing GRMHD simulation results between BAM and SACRA$_{\rm KK22}$., Comment: 21 pages, 21 figures

We investigate the quantization of the Kramers-Kronig (KK) detection technique, initially developed for classical coherent communications. This detection method involves combining the state of interest with a local oscillator on an unbalanced beamsplitter, followed by direct detection and digital signal processing. Intensity measurements in the case of spectrally engineered fields allow for the "digital measurement" of the phase of classical and quantum fields. We show that, to the first order of the local oscillator's amplitude, KK detection is a Gaussian measurement that allows for the estimation of both quadratures of the electric field, similar to double homodyne detection. We study in detail how KK detection operates in the case of bosonic coherent states, pure single-mode and mixed states, as well as the nature of the phase information it measures. Finally, we propose an alternative spectral tomography technique for single-photon states inspired by KK detection.

Pedestrian crowds encompass a complex interplay of intentional movements aimed at reaching specific destinations, fluctuations due to personal and interpersonal variability, and interactions with each other and the environment. Previous work showed the effectiveness of Langevin-like equations in capturing the statistical properties of pedestrian dynamics in simple settings, such as almost straight trajectories. However, modeling more complex dynamics, e.g. when multiple routes and origin-destinations are involved, remains a significant challenge. In this work, we introduce a novel and generic framework to describe the dynamics of pedestrians in any geometric setting, significantly extending previous works. Our model is based on Langevin dynamics with two timescales. The fast timescale corresponds to the stochastic fluctuations present when a pedestrian is walking. The slow timescale is associated with the dynamics that a pedestrian plans to follow, thus a smoother path. Employing a data-driven approach inspired by statistical field theories, we learn the complex potentials directly from the data, namely a high-statistics database of real-life pedestrian trajectories. This approach makes the model generic as the potentials can be read from any trajectory data set and the underlying Langevin structure enables physics-based insights. We validate our model through a comprehensive statistical analysis, comparing simulated trajectories with actual pedestrian measurements across five complementary settings, including a real-life train platform scenario, underscoring its practical societal relevance. We show that our model effectively captures fluctuation statistics in pedestrian motion. Beyond providing fundamental insights and predictive capabilities in pedestrian dynamics, our model could be used to investigate generic active dynamics such as vehicular traffic and collective animal behavior.

This paper studies the power properties of confidence intervals (CIs) for a partially-identified parameter of interest with an interval identified set. We assume the researcher has bounds estimators to construct the CIs proposed by Stoye (2009), referred to as CI1, CI2, and CI3. We also assume that these estimators are "ordered": the lower bound estimator is less than or equal to the upper bound estimator. Under these conditions, we establish two results. First, we show that CI1 and CI2 are equally powerful, and both dominate CI3. Second, we consider a favorable situation in which there are two possible bounds estimators to construct these CIs, and one is more efficient than the other. One would expect that the more efficient bounds estimator yields more powerful inference. We prove that this desirable result holds for CI1 and CI2, but not necessarily for CI3., Comment: 37 pages

We analyze a long-lived hyper-massive neutron star merger remnant (post-merger lifetime $>250$ ms) that has been obtained via large eddy simulations with a gradient subgrid-scale model. We find a clear helicoidal magnetic field structure that is governed by the toroidal component of the magnetic field. Although no jet emerges during the simulation time, we observe at late times a significant increase of the poloidal component of the magnetic field at all scales. We also compare with the results of several binary neutron star simulations with moderate resolution of $120$~m, that are evolved up to $50$~ms after the merger, which differ in terms of the initial topology and strength of the magnetic field. We find that the best choice is an isotropic small-scale magnetic field distribution that mimics the turbulent state that generically develops during the merger. This initial configuration reaches a closer agreement with our high-resolution simulation results than the purely dipolar large-scale fields that are commonly employed in these type of simulations. This provides a recipe to perform such simulations avoiding the computationally expensive grids required to faithfully capture the amplification of the magnetic field by Kelvin-Helmholtz instabilities., Comment: 13 pages, 11 figures

Discerning between authentic content and that generated by advanced AI methods has become increasingly challenging. While previous research primarily addresses the detection of fake faces, the identification of generated natural images has only recently surfaced. This prompted the recent exploration of solutions that employ foundation vision-and-language models, like CLIP. However, the CLIP embedding space is optimized for global image-to-text alignment and is not inherently designed for deepfake detection, neglecting the potential benefits of tailored training and local image features. In this study, we propose CoDE (Contrastive Deepfake Embeddings), a novel embedding space specifically designed for deepfake detection. CoDE is trained via contrastive learning by additionally enforcing global-local similarities. To sustain the training of our model, we generate a comprehensive dataset that focuses on images generated by diffusion models and encompasses a collection of 9.2 million images produced by using four different generators. Experimental results demonstrate that CoDE achieves state-of-the-art accuracy on the newly collected dataset, while also showing excellent generalization capabilities to unseen image generators. Our source code, trained models, and collected dataset are publicly available at: https://github.com/aimagelab/CoDE., Comment: ECCV 2024

The task of sampling efficiently the Gibbs-Boltzmann distribution of disordered systems is important both for the theoretical understanding of these models and for the solution of practical optimization problems. Unfortunately, this task is known to be hard, especially for spin glasses at low temperatures. Recently, many attempts have been made to tackle the problem by mixing classical Monte Carlo schemes with newly devised Neural Networks that learn to propose smart moves. In this article we introduce the Nearest-Neighbours Neural Network (4N) architecture, a physically-interpretable deep architecture whose number of parameters scales linearly with the size of the system and that can be applied to a large variety of topologies. We show that the 4N architecture can accurately learn the Gibbs-Boltzmann distribution for the two-dimensional Edwards-Anderson model, and specifically for some of its most difficult instances. In particular, it captures properties such as the energy, the correlation function and the overlap probability distribution. Finally, we show that the 4N performance increases with the number of layers, in a way that clearly connects to the correlation length of the system, thus providing a simple and interpretable criterion to choose the optimal depth., Comment: 10 pages, 6 figures; SI: 3 pages

Scanning probe microscopy with multi-qubit sensors offers the potential to improve imaging speed and measure previously inaccessible quantities, such as two-point correlations. We develop a multiplexed quantum sensing approach with scanning probes containing two nitrogen-vacancy (NV) centers at the tip apex. A shared optical channel is used for simultaneous qubit initialization and readout, while phase- and frequency-dependent microwave spin manipulations are leveraged for de-multiplexing the optical readout signal. Scanning dual-NV magnetometry is first demonstrated by simultaneously imaging multiple field projections of a ferrimagnetic racetrack device. Then, we record the two-point covariance of spatially correlated field fluctuations across a current-carrying wire. Our multiplex framework establishes a method to investigate a variety of spatio-temporal correlations, including phase transitions and electronic noise, with nanoscale resolution., Comment: 7 pages, 4 figures

The Lov\'asz theta function $\theta(G)$ provides a very good upper bound on the stability number of a graph $G$. It can be computed in polynomial time by solving a semidefinite program (SDP), which also turns out to be fairly tractable in practice. Consequently, $\theta(G)$ achieves a hard-to-beat trade-off between computational effort and strength of the bound. Indeed, several attempts to improve the theta bound are documented, mainly based on playing around the application of the $N_+(\cdot)$ lifting operator of Lov\'asz and Schrijver to the classical formulation of the maximum stable set problem. Experience shows that solving such SDP-s often struggles against practical intractability and requires highly specialized methods. We investigate the application of such an operator to two different linear formulations based on clique and nodal inequalities, respectively. Fewer inequalities describe these two and yet guarantee that the resulting SDP bound is at least as strong as $\theta(G)$. Our computational experience, including larger graphs than those previously documented, shows that upper bounds stronger than $\theta(G)$ can be accessed by a reasonable additional effort using the clique-based formulation on sparse graphs and the nodal-based one on dense graphs.

We report on the fabrication of a Hybrid SAW/BAW resonator made of a thin layer of Sc-doped AlN (AlScN) with a Sc concentration of 40 at% on a 4H-SiC substrate. A Sezawa mode, excited by a vertical electric field, exploits the d31 piezoelectric coefficient to propagate a longitudinal acoustic wave in the AlScN. The resonant frequency is determined via the pitch in the interdigitated transducer (IDT) defined by Deep Ultraviolet (DUV) lithography. The resonant mode travels in the piezoelectric layer without leaking in the substrate thanks to the mismatch in acoustic phase velocities between the piezoelectric and substrate materials. We show the impact of the piezoelectric and IDT layers' thickness on the two found modes. Importantly, we show how thin piezoelectric and electrode layers effectively suppress the Rayleigh mode. While some challenges in the deposition of AlScN remain towards a large coupling coefficient k_eff^2, We show how wave confinement in the IDT obtains a good quality factor. We also show how modifying the IDT reflectivity allows us to engineer a stopband to prevent unwanted modes from being excited between resonance and antiresonance frequencies. Finally, we validate the simulation with fabricated and measured devices and present possible improvements to this resonator architecture., Comment: Submitted to IEEE EDL on 26.07.2024