Search

Decoding and expressing brain activity in a comprehensible form is a challenging frontier in AI. This paper presents Thought2Text, which uses instruction-tuned Large Language Models (LLMs) fine-tuned with EEG data to achieve this goal. The approach involves three stages: (1) training an EEG encoder for visual feature extraction, (2) fine-tuning LLMs on image and text data, enabling multimodal description generation, and (3) further fine-tuning on EEG embeddings to generate text directly from EEG during inference. Experiments on a public EEG dataset collected for six subjects with image stimuli demonstrate the efficacy of multimodal LLMs (LLaMa-v3, Mistral-v0.3, Qwen2.5), validated using traditional language generation evaluation metrics, GPT-4 based assessments, and evaluations by human expert. This approach marks a significant advancement towards portable, low-cost "thoughts-to-text" technology with potential applications in both neuroscience and natural language processing (NLP).

Implicit Neural Representations (INRs) have recently gained attention as a powerful approach for continuously representing signals such as images, videos, and 3D shapes using multilayer perceptrons (MLPs). However, MLPs are known to exhibit a low-frequency bias, limiting their ability to capture high-frequency details accurately. This limitation is typically addressed by incorporating high-frequency input embeddings or specialized activation layers. In this work, we demonstrate that these embeddings and activations are often configured with hyperparameters that perform well on average but are suboptimal for specific input signals under consideration, necessitating a costly grid search to identify optimal settings. Our key observation is that the initial frequency spectrum of an untrained model's output correlates strongly with the model's eventual performance on a given target signal. Leveraging this insight, we propose frequency shifting (or FreSh), a method that selects embedding hyperparameters to align the frequency spectrum of the model's initial output with that of the target signal. We show that this simple initialization technique improves performance across various neural representation methods and tasks, achieving results comparable to extensive hyperparameter sweeps but with only marginal computational overhead compared to training a single model with default hyperparameters., Comment: Code at https://github.com/gmum/FreSh/

Random forests utilize bootstrap sampling to create an individual training set for each component tree. This involves sampling with replacement, with the number of instances equal to the size of the original training set ($N$). Research literature indicates that drawing fewer than $N$ observations can also yield satisfactory results. The ratio of the number of observations in each bootstrap sample to the total number of training instances is called the bootstrap rate (BR). Sampling more than $N$ observations (BR $>$ 1) has been explored in the literature only to a limited extent and has generally proven ineffective. In this paper, we re-examine this approach using 36 diverse datasets and consider BR values ranging from 1.2 to 5.0. Contrary to previous findings, we show that such parameterization can result in statistically significant improvements in classification accuracy compared to standard settings (BR $\leq$ 1). Furthermore, we investigate what the optimal BR depends on and conclude that it is more a property of the dataset than a dependence on the random forest hyperparameters. Finally, we develop a binary classifier to predict whether the optimal BR is $\leq$ 1 or $>$ 1 for a given dataset, achieving between 81.88\% and 88.81\% accuracy, depending on the experiment configuration.

In various scenarios motivated by real life, such as medical data analysis, autonomous driving, and adversarial training, we are interested in robust deep networks. A network is robust when a relatively small perturbation of the input cannot lead to drastic changes in output (like change of class, etc.). This falls under the broader scope field of Neural Network Certification (NNC). Two crucial problems in NNC are of profound interest to the scientific community: how to calculate the robustness of a given pre-trained network and how to construct robust networks. The common approach to constructing robust networks is Interval Bound Propagation (IBP). This paper demonstrates that IBP is sub-optimal in the first case due to its susceptibility to the wrapping effect. Even for linear activation, IBP gives strongly sub-optimal bounds. Consequently, one should use strategies immune to the wrapping effect to obtain bounds close to optimal ones. We adapt two classical approaches dedicated to strict computations -- Dubleton Arithmetic and Affine Arithmetic -- to mitigate the wrapping effect in neural networks. These techniques yield precise results for networks with linear activation functions, thus resisting the wrapping effect. As a result, we achieve bounds significantly closer to the optimal level than IBPs.

We study equivariant wave maps from the $2+1$ dimensional wormhole to the 2-sphere. This model has explicit harmonic map solutions which, in suitable coordinates, have the form of the sine-Gordon kinks/anti-kinks. We conjecture that there exist asymptotically static chains of $N\geq 2$ alternating kinks and anti-kinks whose subsequent rates of expansion increase in geometric progression as $t\rightarrow \infty$. Our argument employs the method of collective coordinates to derive effective finite-dimensional ODE models for the asymptotic dynamics of $N$-chains. For $N=2,3$ the predictions of these effective models are verified by direct PDE computations which demonstrate that the $N$-chains lie at the threshold of kink-anti-kink annihilation., Comment: 16 pages, 6 figures

Advanced microscopy techniques are essential for visualizing and tracking cellular components and molecules in biomedical research. However, conventional fluorescence microscopy methods often struggle with accurately measuring molecule concentrations in cells. To overcome these limitations, we introduce a novel approach that integrates laser scanning confocal microscopy with time-correlated single photon counting (TCSPC), supported by an open-source analysis tool called smICA (single-molecule Image to Concentration Analyzer). Our method, validated against traditional fluorescence correlation spectroscopy (FCS), offers enhanced accuracy in determining fluorescent molecule concentrations, particularly in cases where molecules are immobile or unevenly distributed. This is demonstrated using fluorescently labeled mRNA in living cells, highlighting the approach's effectiveness., Comment: 14 pages, 6 figures, 18, references

We study the propagation of a domain wall (kink) of the $\phi^4$ model in a radially symmetric environment defined by a gravity source. This source deforms the standard Euclidian metric into a Schwartzschild-like one. We introduce an effective model that accurately describes the dynamics of the kink center. This description works well even outside the perturbation region, i.e., even for large masses of the gravitating object. We observed that such a spherical domain wall surrounding a star-type object inevitably "collapses", i.e., shrinks in radius towards the origin and offer an understanding of the latter phenomenology. The relevant analysis is presented for a circular domain wall and a spherical one., Comment: 19 pages, 12 figures

We identify "stable regions" in the residual stream of Transformers, where the model's output remains insensitive to small activation changes, but exhibits high sensitivity at region boundaries. These regions emerge during training and become more defined as training progresses or model size increases. The regions appear to be much larger than previously studied polytopes. Our analysis suggests that these stable regions align with semantic distinctions, where similar prompts cluster within regions, and activations from the same region lead to similar next token predictions. This work provides a promising research direction for understanding the complexity of neural networks, shedding light on training dynamics, and advancing interpretability.

Layered transition metal dichalcogenides (TMDs) stabilize in multiple structural forms with profoundly distinct and exotic electronic phases. Interfacing different layer types is a promising route to manipulate TMDs' properties, not only as a means to engineer quantum devices, but also as a route to explore fundamental physics in complex matter. Here we use angle-resolved photoemission (ARPES) to investigate a strong layering-dependent enhancement of superconductivity in TaS$_2$, in which the superconducting transition temperature, $T_c$, of its $2H$ structural phase is nearly tripled when insulating $1T$ layers are inserted into the system. The study is facilitated by a novel vanadium-intercalation approach to synthesizing various TaS$_2$ polytypes, which improves the quality of ARPES data while leaving key aspects of the electronic structure and properties intact. The spectra show the clear opening of the charge density wave gap in the pure $2H$ phase and its suppression when $1T$ layers are introduced to the system. Moreover, in the mixed-layer $4Hb$ system, we observe a strongly momentum-anisotropic increase in electron-phonon coupling near the Fermi level relative to the $2H$ phase. Both phenomena help to account for the increased $T_c$ in mixed $2H$/$1T$ layer structures.

A mathematical model for the poroelastic materials (PEM) with the variable volume is developed in multidimensional case. Governing equations of the model are constructed using the continuity equations, which reflect the well-known physical laws. The deformation vector is specified using the Terzaghi effective stress tensor. In the two-dimensional space case, the model is studied by analytical methods. Using the classical Lie method, it is proved that the relevant nonlinear system of the (1+2)-dimensional governing equations admits highly nontrivial Lie symmetries leading to an infinite-dimensional Lie algebra. The radially-symmetric case is studied in details. It is shown how correct boundary conditions in the case of PEM in the form of a ring and an annulus are constructed. As a result, boundary-value problems with a moving boundary describing the ring (annulus) deformation are constructed. The relevant nonlinear boundary-value problems are analytically solved in the stationary case. In particular, the analytical formulae for unknown deformations and an unknown radius of the annulus are presented., Comment: 19 pages, 2 figures

We theoretically investigate the effect of ``glancing" collisions on the ultra-high vacuum (UHV) pressure readings of the cold atom vacuum standard (CAVS), based on either ultracold $^7$Li or $^{87}$Rb atoms. Here, glancing collisions are those collisions between ultracold atoms and room-temperature background atoms or molecules in the vacuum that do not impart enough kinetic energy to eject an ultracold atom from its trap. Our model is wholly probabilistic and shows that the number of the ultracold atoms remaining in the trap as a function of time is non-exponential. We update the recent results of a comparison between a traditional pressure standard -- a combined flowmeter and dynamic expansion system -- to the CAVS [D.S. Barker, et al., arXiv:2302.12143] to reflect the results of our model. We find that the effect of glancing collisions shifts the theoretical predictions of the total loss rate coefficients for $^7$Li colliding with noble gases or N$_2$ by up to $0.6$ %. Likewise, we find that in the limit of zero trap depth the experimentally extracted loss rate coefficients for $^{87}$Rb colliding with noble gases or N$_2$ shift by as much as 2.2 %., Comment: 10 pages, 2 figures, 3 tables, and 1 supplemental material with 5 tables

Understanding the decisions made by image classification networks is a critical area of research in deep learning. This task is traditionally divided into two distinct approaches: post-hoc methods and intrinsic methods. Post-hoc methods, such as GradCam, aim to interpret the decisions of pre-trained models by identifying regions of the image where the network focuses its attention. However, these methods provide only a high-level overview, making it difficult to fully understand the network's decision-making process. Conversely, intrinsic methods, like prototypical parts models, offer a more detailed understanding of network predictions but are constrained by specific architectures, training methods, and datasets. In this paper, we introduce InfoDisent, a hybrid model that combines the advantages of both approaches. By utilizing an information bottleneck, InfoDisent disentangles the information in the final layer of a pre-trained deep network, enabling the breakdown of classification decisions into basic, understandable atomic components. Unlike standard prototypical parts approaches, InfoDisent can interpret the decisions of pre-trained classification networks and be used for making classification decisions, similar to intrinsic models. We validate the effectiveness of InfoDisent on benchmark datasets such as ImageNet, CUB-200-2011, Stanford Cars, and Stanford Dogs for both convolutional and transformer backbones.

In this paper, we present two computational methods for performing simulations of pollution propagation described by advection-diffusion equations. The first method employs graph grammars to describe the generation process of the computational mesh used in simulations with the meshless solver of the three-dimensional finite element method. The graph transformation rules express the three-dimensional Rivara longest-edge refinement algorithm. This solver is used for an exemplary application: performing three-dimensional simulations of pollution generation by the coal-burning power plant and its propagation in the city of Longyearbyen, the capital of Spitsbergen. The second computational code is based on the Physics Informed Neural Networks method. It is used to calculate the dissipation of the pollution along the valley in which the city of Longyearbyen is located. We discuss the instantiation and execution of the PINN method using Google Colab implementation. We discuss the benefits and limitations of the PINN implementation., Comment: 34 pages, 21 figures, 1 table

The behavior of the kink in the sine-Gordon (sG) model in the presence of periodic inhomogeneity is studied. An ansatz is proposed that allows for the construction of a reliable effective model with two degrees of freedom. Effective models with excellent agreement with the original field-theoretic partial differential equation are constructed, including in the non-perturbative region and for relativistic velocities. The numerical solutions of the sG model describing the evolution of the kink in the presence of a barrier as well as in the case of a periodic heterogeneity under the potential additional influence of a switched bias current and/or dissipation were obtained. The results of the field equation and the effective models were compared. The effect of the choice of initial conditions in the field model on the agreement of the results with the effective model is discussed., Comment: 21 pages, 17 figures

Modern e-commerce platforms offer vast product selections, making it difficult for customers to find items that they like and that are relevant to their current session intent. This is why it is key for e-commerce platforms to have near real-time scalable and adaptable personalized ranking and search systems. While numerous methods exist in the scientific literature for building such systems, many are unsuitable for large-scale industrial use due to complexity and performance limitations. Consequently, industrial ranking systems often resort to computationally efficient yet simplistic retrieval or candidate generation approaches, which overlook near real-time and heterogeneous customer signals, which results in a less personalized and relevant experience. Moreover, related customer experiences are served by completely different systems, which increases complexity, maintenance, and inconsistent experiences. In this paper, we present a personalized, adaptable near real-time ranking platform that is reusable across various use cases, such as browsing and search, and that is able to cater to millions of items and customers under heavy load (thousands of requests per second). We employ transformer-based models through different ranking layers which can learn complex behavior patterns directly from customer action sequences while being able to incorporate temporal (e.g. in-session) and contextual information. We validate our system through a series of comprehensive offline and online real-world experiments at a large online e-commerce platform, and we demonstrate its superiority when compared to existing systems, both in terms of customer experience as well as in net revenue. Finally, we share the lessons learned from building a comprehensive, modern ranking platform for use in a large-scale e-commerce environment.

Real-world processes often exhibit temporal separation between actions and reactions - a characteristic frequently ignored in many modelling frameworks. Adding temporal aspects, like time delays, introduces a higher complexity of problems and leads to models that are challenging to analyse and computationally expensive to solve. In this work, we propose an intermediate solution to resolve the issue in the framework of evolutionary game theory. Our compartment-based model includes time delays while remaining relatively simple and straightforward to analyse. We show that this model yields qualitatively comparable results with models incorporating explicit delays. Particularly, we focus on the case of delays between parents' interaction and an offspring joining the population, with the magnitude of the delay depending on the parents' strategy. We analyse Stag-Hunt, Snowdrift, and the Prisoner's Dilemma game and show that strategy-dependent delays are detrimental to affected strategies. Additionally, we present how including delays may change the effective games played in the population, subsequently emphasising the importance of considering the studied systems' temporal aspects to model them accurately.

Knowledge probing evaluates the extent to which a language model (LM) has acquired relational knowledge during its pre-training phase. It provides a cost-effective means of comparing LMs of different sizes and training setups and is useful for monitoring knowledge gained or lost during continual learning (CL). In prior work, we presented an improved knowledge probe called BEAR (Wiland et al., 2024), which enables the comparison of LMs trained with different pre-training objectives (causal and masked LMs) and addresses issues of skewed distributions in previous probes to deliver a more unbiased reading of LM knowledge. With this paper, we present LM-PUB- QUIZ, a Python framework and leaderboard built around the BEAR probing mechanism that enables researchers and practitioners to apply it in their work. It provides options for standalone evaluation and direct integration into the widely-used training pipeline of the Hugging Face TRANSFORMERS library. Further, it provides a fine-grained analysis of different knowledge types to assist users in better understanding the knowledge in each evaluated LM. We publicly release LM-PUB-QUIZ as an open-source project.

We introduce and analyze quasi-local mass using Hamiltonian methods. It is based on multipole decomposition for surfaces that are topological spheres. Based on the above model, tests were performed for Kerr spacetime for two arbitrary choices of surfaces: rigid spheres and round spheres., Comment: 35 pages, no figures

This paper presents an experimental study on mmWave beam profiling on a mmWave testbed, and develops a machine learning model for beamforming based on the experiment data. The datasets we have obtained from the beam profiling and the machine learning model for beamforming are valuable for a broad set of network design problems, such as network topology optimization, user equipment association, power allocation, and beam scheduling, in complex and dynamic mmWave networks. We have used two commercial-grade mmWave testbeds with operational frequencies on the 27 Ghz and 71 GHz, respectively, for beam profiling. The obtained datasets were used to train the machine learning model to estimate the received downlink signal power, and data rate at the receivers (user equipment with different geographical locations in the range of a transmitter (base station). The results have shown high prediction accuracy with low mean square error (loss), indicating the model's ability to estimate the received signal power or data rate at each individual receiver covered by a beam. The dataset and the machine learning-based beamforming model can assist researchers in optimizing various network design problems for mmWave networks., Comment: In Proceedings of IEEE International Conference on Computer Communications and Networks (ICCCN), 2023

Recognizing errors in assembly and maintenance procedures is valuable for industrial applications, since it can increase worker efficiency and prevent unplanned down-time. Although assembly state recognition is gaining attention, none of the current works investigate assembly error localization. Therefore, we propose StateDiffNet, which localizes assembly errors based on detecting the differences between a (correct) intended assembly state and a test image from a similar viewpoint. StateDiffNet is trained on synthetically generated image pairs, providing full control over the type of meaningful change that should be detected. The proposed approach is the first to correctly localize assembly errors taken from real ego-centric video data for both states and error types that are never presented during training. Furthermore, the deployment of change detection to this industrial application provides valuable insights and considerations into the mechanisms of state-of-the-art change detection algorithms. The code and data generation pipeline are publicly available at: https://timschoonbeek.github.io/error_seg., Comment: 23 pages (14 main paper, 2 references, 7 supplementary), 15 figures (8 main paper, 7 supplementary). Accepted at ECCV Vision-based InduStrial InspectiON (VISION) workshop

Let $\{P_t\}_{t>0}$ be the Dunkl-Poisson semigroup associated with a root system $R\subset \mathbb R^N$ and a multiplicity function $k\geq 0$. Analogously to the classical theory, we say that a bounded measurable function $f$ defined on $\mathbb R^N$ belongs to the inhomogeneous Lipschitz space $\Lambda_k^\beta$, $\beta>0$, if $$\sup_{t>0} t^{m-\beta} \Big\|\frac{d^m}{dt^m} P_tf\Big\|_{L^\infty}<\infty,$$ where $m=[\beta]+1$. We prove that the spaces $\Lambda^\beta_k$ coincide with the classical Lipschitz spaces., Comment: 32 pages

Assembly state recognition facilitates the execution of assembly procedures, offering feedback to enhance efficiency and minimize errors. However, recognizing assembly states poses challenges in scalability, since parts are frequently updated, and the robustness to execution errors remains underexplored. To address these challenges, this paper proposes an approach based on representation learning and the novel intermediate-state informed loss function modification (ISIL). ISIL leverages unlabeled transitions between states and demonstrates significant improvements in clustering and classification performance for all tested architectures and losses. Despite being trained exclusively on images without execution errors, thorough analysis on error states demonstrates that our approach accurately distinguishes between correct states and states with various types of execution errors. The integration of the proposed algorithm can offer meaningful assistance to workers and mitigate unexpected losses due to procedural mishaps in industrial settings. The code is available at: https://timschoonbeek.github.io/state_rec, Comment: 8 pages, 8 figures

Exploring the non-Hermitian properties of semiconductor materials for optical applications is at the forefront of photonic research. However, the selection of appropriate systems to implement such photonic devices remains a topic of debate. In this work, we demonstrate that a perovskite crystal, characterized by its easy and low-cost manufacturing, when placed between two distributed Bragg reflectors with an air gap, can form a natural double microcavity. This construction shows promising properties for the realisation of novel, tunable non-Hermitian photonic devices through strong light-matter coupling. We reveal that such a system exhibits double-coupled polariton modes with dispersion including multiple inflection points. Owing to its non-Hermiticity, our system exhibits nonreciprocal properties and allows for the observation of exceptional points. Our experimental studies are in agreement with the theoretical analysis based on coupled mode theory and calculations based on transfer matrix method., Comment: Accepted manuscript (AM)

In this paper the relations between independence preserving (IP) involutions and reversible Markov kernels are investigated. We introduce an involutive augmentation H = (f, g_f) of a measurable function f and relate the IP property of H to f-generated reversible Markov kernels. Various examples appeared in the literature are presented as particular cases of the construction. In particular, we prove that the IP property generated by the (reversible) Markov kernel of random walk with a reflecting barrier at the origin characterizes geometric-type laws, Comment: 14 pages

The tunability of individual coupling amplitudes in photonic lattices is highly desirable for photonic Hamiltonian engineering and for studying topological transitions in situ. In this work, we demonstrate the tunneling control between individual lattice sites patterned inside an optical microcavity. The tuning is achieved by applying a voltage to a liquid crystal microcavity possessing photonic Rashba-Dresselhaus spin-orbit coupling. This type of spin-orbit coupling emerges due to the high birefringence of the liquid crystal material and constitutes an artificial gauge field for photons. The proposed technique can be combined with strong-light matter coupling and non-Hermitian physics already established in liquid crystal microcavities.

Theoretical considerations are made for the nuclear excitation from the ground state to the 8 eV $^{229m}$Th isomer via near-resonant electron transitions in Sb-like ($q=39+$) thorium ions. The energy of the first excited atomic state ($J=7/2$) in the $^{229}$Th$^{39+}$ ion is estimated to be 8.308$\pm$0.011 eV, which is very close to the new reference value for the $^{229m}$Th nuclear isomer energy, 8.338$\pm$0.024 eV [Kraemer et al., Nature 617, 706 (2023)]. It was found that within the uncertainty range of both atomic and nuclear excitation energies, the rate of nuclear excitation by electron transition can vary by more than 20 orders of magnitude. Our results indicate that the upper theoretical limit for the $^{229m}$Th isomer excitation rate reaches an enormous value of $1.05\times10^{16}$ s$^{-1}$ at resonance ($\Delta=0$ meV). Additionally, it was shown that using an electron beam ion trap (EBIT), the production of the $^{229}$Th isomer can reach rates ranging from tenths to approximately $6\times10^{19}$ s$^{-1}$. Thus, EBIT experimental studies could serve as an extremely sensitive tool for the independent verification of the excitation energy of the isomeric state in the $^{229}$Th nucleus.

We present an extensive review of the two-dimensional finite difference Hartree--Fock (FD HF) method, and present its implementation in the newest version of X2DHF, the FD HF program for atoms and diatomic molecules. The program was originally published in Comput. Phys. Commun. in 1996, and was last revised in 2013. X2DHF can be used to obtain HF limit values of total energies and multipole moments for a wide range of diatomic molecules and their ions, using either point nuclei or a finite nuclear model. Polarizabilities ($\alpha_{zz}$) and hyperpolarizabilities ($\beta_{zzz}$, $\gamma_{zzzz}$, ${A}_{z,zz}$, ${B}_{zz,zz}$) can also be computed by the program with the finite-field method. X2DHF has been extensively used in the literature to assess the accuracy of existing atomic basis sets and to help in developing new ones. As a new feature since the last revision, the program can now also perform Kohn-Sham density functional calculations with local and generalized gradient exchange-correlation functionals with the Libxc library of density functionals, enabling new types of studies. Furthermore, the initialization of calculations has been greatly simplified. As before, X2DHF can also perform one-particle calculations with (smooth) Coulomb, Green-Sellin-Zachor and Krammers-Henneberger potentials, while calculations with a superposition of atomic potentials have been added as a new feature. The program is easy to install from the GitHub repository and build via CMake using the x2dhfctl script that facilitates creating its single- and multiple-threaded versions, as well as building in Libxc support. Calculations can be carried out with X2DHF in double- or quadruple-precision arithmetic., Comment: 116 pages, 14 figures

Recent years have seen a tremendous improvement in the quality of video generation and editing approaches. While several techniques focus on editing appearance, few address motion. Current approaches using text, trajectories, or bounding boxes are limited to simple motions, so we specify motions with a single motion reference video instead. We further propose to use a pre-trained image-to-video model rather than a text-to-video model. This approach allows us to preserve the exact appearance and position of a target object or scene and helps disentangle appearance from motion. Our method, called motion-textual inversion, leverages our observation that image-to-video models extract appearance mainly from the (latent) image input, while the text/image embedding injected via cross-attention predominantly controls motion. We thus represent motion using text/image embedding tokens. By operating on an inflated motion-text embedding containing multiple text/image embedding tokens per frame, we achieve a high temporal motion granularity. Once optimized on the motion reference video, this embedding can be applied to various target images to generate videos with semantically similar motions. Our approach does not require spatial alignment between the motion reference video and target image, generalizes across various domains, and can be applied to various tasks such as full-body and face reenactment, as well as controlling the motion of inanimate objects and the camera. We empirically demonstrate the effectiveness of our method in the semantic video motion transfer task, significantly outperforming existing methods in this context., Comment: Preprint. All videos in this paper are best viewed as animations with Acrobat Reader by pressing the highlighted frame of each video

We present a unified approach to two classes of Banach spaces defined with the aid of variations: Waterman spaces and Chanturia classes. Our method is based on some ideas coming from the theory of ideals on the set of natural numbers.

Unusual magnetic properties of Van der Waals type antiferromagnetic semiconductors make them highly attractive for spintronics and optoelectronics. A link between the magnetic and optical properties of those materials, required for practical applications, has not been, however, established so far. Here, we report on a combined experimental and theoretical study of magnetic, optical, and structural properties of bulk CrPS$_{4}$ samples. We find that the magnetic-field-dependent circular polarization degree of the photoluminescence is a direct measure of the net magnetization of CrPS$_{4}$. Complementary, Raman scattering measured as a function of magnetic field and temperature enables the determination of the magnetic susceptibility curve of the material. Our experimental results are backed by Our experimental results are supported by density functional theory calculations that take as input the lattice parameters determined from temperature-dependent X-ray diffraction measurements. This allows us to explain the impact of spin ordering on the spectral position of Raman transitions in CrPS$_4$, as well as anomalous temperature shifts of selected of them. The presented method for all-optical determination of the magnetic properties is highly promising for studies of spin ordering and magnetic phase transitions in single- or a few-layer samples of magnetic layered materials, for which a poor signal-to-noise ratio precludes any reliable neutron scattering or magnetometry measurements.

Diffusion models are among the most effective methods for image generation. This is in particular because, unlike GANs, they can be easily conditioned during training to produce elements with desired class or properties. However, guiding a pre-trained diffusion model to generate elements from previously unlabeled data is significantly more challenging. One of the possible solutions was given by the ADM-G guiding approach. Although ADM-G successfully generates elements from the given class, there is a significant quality gap compared to a model originally conditioned on this class. In particular, the FID score obtained by the ADM-G-guided diffusion model is nearly three times lower than the class-conditioned guidance. We demonstrate that this issue is partly due to ADM-G providing minimal guidance during the final stage of the denoising process. To address this problem, we propose GeoGuide, a guidance model based on tracing the distance of the diffusion model's trajectory from the data manifold. The main idea of GeoGuide is to produce normalized adjustments during the backward denoising process. As shown in the experiments, GeoGuide surpasses the probabilistic approach ADM-G with respect to both the FID scores and the quality of the generated images.

At its core, information access and seeking is an interactive process. In existing search engines, interactions are limited to a few pre-defined actions, such as "requery", "click on a document", "scrolling up/down", "going to the next result page", "leaving the search engine", etc. A major benefit of moving towards generative IR systems is enabling users with a richer expression of information need and feedback and free-form interactions in natural language and beyond. In other words, the actions users take are no longer limited by the clickable links and buttons available on the search engine result page and users can express themselves freely through natural language. This can go even beyond natural language, through images, videos, gestures, and sensors using multi-modal generative IR systems. This chapter briefly discusses the role of interaction in generative IR systems. We will first discuss different ways users can express their information needs by interacting with generative IR systems. We then explain how users can provide explicit or implicit feedback to generative IR systems and how they can consume such feedback. Next, we will cover how users interactively can refine retrieval results. We will expand upon mixed-initiative interactions and discuss clarification and preference elicitation in more detail. We then discuss proactive generative IR systems, including context-aware recommendation, following up past conversations, contributing to multi-party conversations, and feedback requests. Providing explanation is another interaction type that we briefly discuss in this chapter. We will also briefly describe multi-modal interactions in generative information retrieval. Finally, we describe emerging frameworks and solutions for user interfaces with generative AI systems., Comment: Draft of a chapter intended to appear in a forthcoming book on generative information retrieval, co-edited by Chirag Shah and Ryen White

Time sharing between activities remains an indispensable part of everyday activity pattern. However, the issue has not yet been fully acknowledged within the existing time allocation models, potentially resulting in inaccuracies in valuing travel time savings. Therefore this study is aimed at addressing this gap by investigating the potential impact of introducing time sharing within such a framework, as well as factors determining it as represented by travel activities. In doing so, time constraint in the time allocation model of Small was modified to enable sharing the same time interval between different activities. The resulting expression indicated that such an augmentation could lead to lower estimates of value of time as a resource. On the other hand, empirical research based on the data from the National Passenger Survey 2004 used for calibrating cross-nested logit model indicated a number of factors affecting the choice of travel activities. It was discovered that significant include possession of equipment allowing particular activities, e.g. newspaper, paperwork or ICT devices, companionship, gender, length of the journey, frequency of using the service, possibility of working on the train, journey planning in advance, first class travel, termination of the trip in central London, peak-time travel and availability of seating., Comment: 15 pages, 2 Figures, Presented at European Transport Conference 2010

We create a one-dimensional photonic crystal with strong polarization dependence and tunable by an applied electric field. We accomplish this in a planar microcavity by embedding a cholesteric liquid crystal (LC), which spontaneously forms a uniform lying helix (ULH). The applied voltage controls the orientation of the LC molecules and, consequently, the strength of a polarization-dependent periodic potential. It leads to opening or closing of photonic band gaps in the dispersion of the massive photons in the microcavity. In addition, when the ULH structure possesses a molecular tilt, it induces a spin-orbit coupling between the lattice bands of different parity. This interband spin-orbit coupling (ISOC) is analogous to optical activity and can be treated as a synthetic non-Abelian gauge potential. Finally, we show that doping the LC with dyes allows us to achieve lasing that inherits all the above-mentioned tunable properties of LC microcavity, including dual and circularly-polarized lasing.

We identify persistent oscillations in a nonintegrable quantum Ising chain left behind by a rapid transition into a ferromagnetic phase. In the integrable chain with nearest-neighbor (NN) interactions, the nature, origin, and decay of post-transition oscillations are tied to the Kibble-Zurek mechanism (KZM). However, when coupling to the next nearest neighbor (NNN) is added, the resulting nonintegrable Ising chain (still in the quantum Ising chain universality class) supports persistent post-transition oscillation: KZM-like oscillations turn into persistent oscillations of transverse magnetization. Their longevity in our simulations is likely limited only by the numerical accuracy. Their period differs from the decaying KZM oscillation but their amplitude depends on quench rate. Moreover, they can be excited by driving in resonance with the excitations' energy gap. Thus, while one might have expected that the integrability-breaking NNN coupling would facilitate relaxation, the oscillations we identify are persistent. At low to medium transverse fields, they are associated with Cooper pairs of Bogoliubov quasiparticles -- kinks. This oscillation of the pair condensate is a manifestation of quantum coherence., Comment: 6 pages, 5 figures, plus Supplementary Material

Despite numerous efforts from numerical approaches to complement experimental measurements, several fundamental challenges have still hindered one's ability to truly provide an atomistic picture of the nucleation process in nanocrystals. Among them, our study resolves three obstacles: (1) Machine-learning force fields including long-range interactions able to capture the finesse of the underlying atomic interactions, (2) Data-driven characterization of the local ordering in a complex structural landscape associated with several crystal polymorphs and (3) Comparing results from a large range of temperatures using both brute-force and rare-event sampling. Altogether, our simulation strategy has allowed us to study zinc oxide crystallization from nano-droplet melt. Remarkably, our results show that different nucleation pathways compete depending on the investigated degree of supercooling.

In Bose-Einstein condensates, spin-orbit coupling produces supersolidity. It is a peculiar state of matter, which, in addition to the superfluid behaviour of weakly interacting Bose condensates, shows a periodic modulation of its density typical for crystals and called stripe phase. Here, we report the fabrication of a new type of samples allowing to achieve room-temperature supersolidity for a quantum fluid of light. The structure is an optical microcavity filled with a nematic liquid crystal (LC) sandwiched between two layers of the organic polymer MeLPPP. We demonstrate the formation of cavity exciton-polaritons in the presence of Rashba-Dresselhaus spin-orbit coupling (RDSOC), which is tuned by external voltage controlling the LC birefringence. In the RDSOC regime, we demonstrate exciton-polariton condensation in the two distinct degenerate minima of the dispersion. The condensate real space distribution shows both polarization and density stripes, which stem from the interference between phase-coherent condensate components characterized by different wavevectors and polarizations. The possibilities offered by this platform to tune the particle dispersion and to perform full state tomography, including time-resolved, open wide perspectives for detailed future studies of the static and dynamical behaviour of supersolids and of quantum fluids in presence of SOC and topologically non-trivial bands.

Fundamental properties of spontaneous and piezo polarization are reformulated and critically reviewed. It was demonstrated that Landau definition of polarization as a dipole density could be used to the infinite systems. The difference between the bulk polarization and surface polarity are distinguished thus creating clear identification of both components. The local model of spontaneous polarization was created and used to calculate spontaneous polarization as the electric dipole density. It was shown that the proposed local model correctly predicts c-axis spontaneous polarization values of the nitride wurtzite semiconductors. It was also shown that the proposed model predicts zero polarization in the plane perpendicular to the c-axis, in accordance with symmetry requirements. In addition, the model results are in accordance with polarization equal to zero for zinc blende lattice. These data confirm the basic correctness of the proposed model. The spontaneous polarization values obtained for all wurtzite III nitrides (BN, AlN, GaN and InN) are in basic agreement with the earlier calculations using Berry phase and slab models of Bernardini et al. {Bernardini et al. Phys Rev B 56 (2001) R10024 & 63 (2001) 193201} but not with Dreyer et al. {Dreyer et al. Phys. Rev X 6 (2016) 021038}. Wurtzite nitride superlattices ab initio calculations were performed to derive polarization-induced fields in the coherently strained lattices showing good agreement with the polarization values. The strained superlattice data were used to determine the piezoelectric parameters of wurtzite nitrides obtaining the values that were in basic agreement with the earlier data. Zinc blende superlattices were also modeled using ab initio calculations showing results that are in agreement with the absence of polarization of all nitrides in zinc blende symmetry., Comment: 36 pages, 13 figures, 53 references

A project (e.g., writing a collaborative research paper) is often a group effort. At the end, each contributor identifies their contribution, often verbally. The reward, however, is very frequently financial. It leads to the question of what (percentage) share in the creation of the paper is due to individual authors. Different authors may have various opinions on the matter; even worse, their opinions may have different relevance. In this paper, we present simple models that allow aggregation of experts' views, linking the priority of his preference directly to the assessment made by other experts. In this approach, the more significant the contribution of a given expert, the greater the importance of his opinion. The presented method can be considered an attempt to find consensus among peers involved in the same project. Hence, its applications may go beyond the proposed study example of writing a scientific paper., Comment: 25 pages, 3 figures

Technology acceptance models effectively predict how users will adopt new technology products. Traditional surveys, often expensive and cumbersome, are commonly used for this assessment. As an alternative to surveys, we explore the use of large language models for annotating online user-generated content, like digital reviews and comments. Our research involved designing an LLM annotation system that transform reviews into structured data based on the Unified Theory of Acceptance and Use of Technology model. We conducted two studies to validate the consistency and accuracy of the annotations. Results showed moderate-to-strong consistency of LLM annotation systems, improving further by lowering the model temperature. LLM annotations achieved close agreement with human expert annotations and outperformed the agreement between experts for UTAUT variables. These results suggest that LLMs can be an effective tool for analyzing user sentiment, offering a practical alternative to traditional survey methods and enabling deeper insights into technology design and adoption., Comment: This is a preprint of a paper accepted for the 32nd International Conference on Information Systems Development (ISD 2024), Gdansk, Poland

Single-step retrosynthesis aims to predict a set of reactions that lead to the creation of a target molecule, which is a crucial task in molecular discovery. Although a target molecule can often be synthesized with multiple different reactions, it is not clear how to verify the feasibility of a reaction, because the available datasets cover only a tiny fraction of the possible solutions. Consequently, the existing models are not encouraged to explore the space of possible reactions sufficiently. In this paper, we propose a novel single-step retrosynthesis model, RetroGFN, that can explore outside the limited dataset and return a diverse set of feasible reactions by leveraging a feasibility proxy model during the training. We show that RetroGFN achieves competitive results on standard top-k accuracy while outperforming existing methods on round-trip accuracy. Moreover, we provide empirical arguments in favor of using round-trip accuracy which expands the notion of feasibility with respect to the standard top-k accuracy metric.

Machine learning has vast potential to improve anomaly detection in satellite telemetry which is a crucial task for spacecraft operations. This potential is currently hampered by a lack of comprehensible benchmarks for multivariate time series anomaly detection, especially for the challenging case of satellite telemetry. The European Space Agency Benchmark for Anomaly Detection in Satellite Telemetry (ESA-ADB) aims to address this challenge and establish a new standard in the domain. It is a result of close cooperation between spacecraft operations engineers from the European Space Agency (ESA) and machine learning experts. The newly introduced ESA Anomalies Dataset contains annotated real-life telemetry from three different ESA missions, out of which two are included in ESA-ADB. Results of typical anomaly detection algorithms assessed in our novel hierarchical evaluation pipeline show that new approaches are necessary to address operators' needs. All elements of ESA-ADB are publicly available to ensure its full reproducibility., Comment: 87 pages, 24 figures, 19 tables

Label noise remains a challenge for training robust classification models. Most methods for mitigating label noise have been benchmarked using primarily datasets with synthetic noise. While the need for datasets with realistic noise distribution has partially been addressed by web-scraped benchmarks such as WebVision and Clothing1M, those benchmarks are restricted to the computer vision domain. With the growing importance of Transformer-based models, it is crucial to establish text classification benchmarks for learning with noisy labels. In this paper, we present AlleNoise, a new curated text classification benchmark dataset with real-world instance-dependent label noise, containing over 500,000 examples across approximately 5,600 classes, complemented with a meaningful, hierarchical taxonomy of categories. The noise distribution comes from actual users of a major e-commerce marketplace, so it realistically reflects the semantics of human mistakes. In addition to the noisy labels, we provide human-verified clean labels, which help to get a deeper insight into the noise distribution, unlike web-scraped datasets typically used in the field. We demonstrate that a representative selection of established methods for learning with noisy labels is inadequate to handle such real-world noise. In addition, we show evidence that these algorithms do not alleviate excessive memorization. As such, with AlleNoise, we set the bar high for the development of label noise methods that can handle real-world label noise in text classification tasks. The code and dataset are available for download at https://github.com/allegro/AlleNoise.

In recent years, epsilon-near-zero (ENZ) materials have attracted much attention due to their unique properties that can be tuned under electrical and optical signals. Furthermore, they allow for a strong enhancement of a nonlinearity close to the ENZ regime, which can have a direct impact on many fields ranging from telecommunications, sensing, quantum optics, to neuromorphic computing. Among the many materials belonging to the ENZ class, transparent conductive oxides (TCOs) are of particular interest because they are very well known to the community and possess very well understood electrical and optical properties. This work extends the direct control of the optical properties of transparent conductive oxides by tailoring the film thickness, which opens new possibilities for ENZ-enhanced photonic applications. It is shown that the thickness-dependent ENZ resonance in the TCO films is equivalent to the power-dependent ENZ resonance as for the same amount of power provided to the TCO film the electric field confinement in the thinner films is highly enhanced for TM polarized light, resulting in the higher electrons effective mass. Thus, the optical properties of TCO can be optically tuned in the ENZ regime, which opens possibilities for a new type of devices that can operate under all-optical switching mechanism., Comment: 10 pages, 4 figures. arXiv admin note: text overlap with arXiv:2309.14454

Adiabatic preparation of a critical ground state is hampered by the closing of its energy gap as the system size increases. However, this gap is directly relevant only for a uniform ramp, where a control parameter in the Hamiltonian is tuned uniformly in space towards the quantum critical point. Here, we consider inhomogeneous ramps in two dimensions: initially, the parameter is made critical at the center of a lattice, from where the critical region expands at a fixed velocity. In the 1D and 2D quantum Ising models, which have a well-defined speed of sound at the critical point, the ramp becomes adiabatic with a subsonic velocity. This subsonic ramp can prepare the critical state faster than a uniform one. Moreover, in both a model of $p$-wave paired 2D fermions and the Kitaev model, the critical dispersion is anisotropic -- linear with a nonzero velocity in one direction and quadratic in the other -- but the gap is still inversely proportional to the linear size of the critical region, with a coefficient proportional to the nonzero velocity. This suffices to make the inhomogeneous ramp adiabatic below a finite crossover velocity and superior to the homogeneous one., Comment: 12 pages, 9 figures; Updated with changes made for publication in Physical Review B

The Extreme Universe Space Observatory on a Super Pressure Balloon 2 (EUSO-SPB2) flew on May 13$^{\text{th}}$ and 14$^{\text{th}}$ of 2023. Consisting of two novel optical telescopes, the payload utilized next-generation instrumentation for the observations of extensive air showers from near space. One instrument, the fluorescence telescope (FT) searched for Ultra-High Energy Cosmic Rays (UHECRs) by recording the atmosphere below the balloon in the near-UV with a 1~$\mu$s time resolution using 108 multi-anode photomultiplier tubes with a total of 6,912 channels. Validated by pre-flight measurements during a field campaign, the energy threshold was estimated around 2~EeV with an expected event rate of approximately 1 event per 10 hours of observation. Based on the limited time afloat, the expected number of UHECR observations throughout the flight is between 0 and 2. Consistent with this expectation, no UHECR candidate events have been found. The majority of events appear to be detector artifacts that were not rejected properly due to a shortened commissioning phase. Despite the earlier-than-expected termination of the flight, data were recorded which provide insights into the detectors stability in the near-space environment as well as the diffuse ultraviolet emissivity of the atmosphere, both of which are impactful to future experiments.

We extend the classical Baire category approach, used in proving the finite generator theorem of Krieger, the homomorphism theorem of Sinai and the isomorphism theorem of Ornstein, applying a similar reasoning to the case of actions of countably infinite amenable groups. In principle we follow the lines of the paper by Burton, Keane and Serafin (\cite{BKS}), showing that measures defining homomorphisms or isomorphisms form residual subsets in suitably chosen spaces of joinings.

We present a new method to solve the nuclear density functional theory (DFT) equations using a two-center harmonic oscillator basis, incorporating pairing and Coulomb interactions. The goal is to efficiently determine the fission and fusion configurations in nuclei. The Coulomb exchange terms are evaluated exactly, allowing for a novel approach to neck formation without the Slater approximation, which has been commonly used in space coordinate-based approaches. The new method has been implemented in the code {\sc hfodd}, enabling direct comparison with standard one-center solutions. This first paper focuses on deriving and implementing a methodology based on stable, precise, and exact applications of harmonic oscillator bases for the two fragments, which can either overlap or be separated by arbitrarily large distances. The implementation is tested on two proof-of-principle examples using light nuclei, specifically $^8$Be and $^{24}$Mg., Comment: Added new reference pertaining CDFT (now ref [24]), two-center shell model (now ref [14,15] and [18]). Updated reference from arXiv to PRC (ref [24]). Modified paragraphs 7th and 8th of the introductory section. Added more references

Video processing is generally divided into two main categories: processing of the entire video, which typically yields optimal classification outcomes, and real-time processing, where the objective is to make a decision as promptly as possible. The latter is often driven by the need to identify rapidly potential critical or dangerous situations. These could include machine failure, traffic accidents, heart problems, or dangerous behavior. Although the models dedicated to the processing of entire videos are typically well-defined and clearly presented in the literature, this is not the case for online processing, where a plethora of hand-devised methods exist. To address this, we present \our{}, a novel, unified, and theoretically-based adaptation framework for dealing with the online classification problem for video data. The initial phase of our study is to establish a robust mathematical foundation for the theory of classification of sequential data, with the potential to make a decision at an early stage. This allows us to construct a natural function that encourages the model to return an outcome much faster. The subsequent phase is to demonstrate a straightforward and readily implementable method for adapting offline models to online and recurrent operations. Finally, by comparing the proposed approach to the non-online state-of-the-art baseline, it is demonstrated that the use of \our{} encourages the network to make earlier classification decisions without compromising accuracy.