Search

This report is the result of the participation of hundreds of students, families, school educators, and researchers who dedicated their time and energy to investigating the critical role that families and communities play in ensuring students and schools can flourish. It is a culmination of over two years of collaborative research and hundreds of conversations on six continents. While there were unique findings in each school, district, and country, six powerful lessons stand out across geographies and contexts. This research report delves into these lessons and how to build greater family, school, and community partnerships as seen through the eyes of families, educators, and students who shared their beliefs on, experiences with, and trust in schools. After venturing into government schools across rural and urban districts in Sierra Leone to facilitate conversations with families and communities, one of the lead researchers noted that there is a crucial and symbiotic relationship between schools, families, and communities that are often overlook.

Chiral active materials are those that break both time-reversal symmetry and parity microscopically, which results in average rotation of the material's complex molecules around their center-of-mass (CM). These materials are far from equilibrium due to their local non-vanishing spin angular momentum. In this paper we show that, unlike passive fluids, the non vanishing spin angular momentum brings about a difference between the CM momentum and the total momentum, which accounts for the momentum of all atoms that compose the complex rotating molecules. This is in stark contrast to equilibrium fluids where the CM stress and the total momentum are essentially equivalent. In fact, we find that generally the CM dynamics are insufficient to describe the dynamics of a chiral active material. The total momentum, other than being experimentally accessible in simple rheological experiments, also imposes another constraint -- its stress must be allowed to be written in a symmetric way. We find that the latter imposes a relation between possible central-force interactions and spin-spin interactions, and constraints the ammount of {\it odd} viscosities in the system to the well-known odd (Hall) viscosity and the odd pressure., Comment: 8 pages, 1 figure

We suggest a novel algorithm for online change point detection based on sequential score function estimation and tracking the best expert approach. The core of the procedure is a version of the fixed share forecaster for the case of infinite number of experts and quadratic loss functions. The algorithm shows a promising performance in numerical experiments on artificial and real-world data sets. We also derive new upper bounds on the dynamic regret of the fixed share forecaster with varying parameter, which are of independent interest., Comment: 43 pages, 4 figures

Materials that are constantly driven out of thermodynamic equilibrium, such as active and living systems, typically violate the Einstein relation. This may arise from active contributions to particle fluctuations which are unrelated to the dissipative resistance of the surrounding medium. We show that in these cases the widely used relation between informatic entropy production and heat dissipation does not hold. Consequently, fluctuation relations for the mechanical work, such as the Jarzynski and Crooks theorems, are invalid. We relate the breaking of the correspondence between entropy production and heat dissipation to departure from the fluctuation-dissipation theorem. We propose a temperaturelike variable that restores this correspondence and gives rise to a generalized second law of thermodynamics, whereby the dissipated heat is necessarily non-negative and vanishes at equilibrium. The Clausius inequality, Carnot maximum efficiency theorem, and relation between the extractable work and the change of free energy are recovered as well., Comment: 15 pages

Modern large language models (LLMs) represent a paradigm shift in what can plausibly be expected of machine learning models. The fact that LLMs can effectively generate sensible answers to a diverse range of queries suggests that they would be useful in customer support applications. While powerful, LLMs have been observed to be prone to hallucination which unfortunately makes their near term use in customer support applications challenging. To address this issue we present a system that allows us to use an LLM to augment our customer support advocates by re-framing the language modeling task as a discriminative classification task. In this framing, we seek to present the top-K best template responses for a customer support advocate to use when responding to a customer. We present the result of both offline and online experiments where we observed offline gains and statistically significant online lifts for our experimental system. Along the way, we present observed scaling curves for validation loss and top-K accuracy, resulted from model parameter ablation studies. We close by discussing the space of trade-offs with respect to model size, latency, and accuracy as well as and suggesting future applications to explore.

Stochastic thermodynamics is a developing theory for systems out of thermal equilibrium. It allows to formulate a wealth of nontrivial relations among thermodynamic quantities such as heat dissipation, excess work, and entropy production in generic nonequilibrium stochastic processes. A key quantity for the derivation of these relations is the propagator - the probability to observe a transition from one point in phase space to another after a given time. Here, applying stochastic Taylor expansions, we devise a formal expansion procedure for the propagator of overdamped Langevin dynamics. The three leading orders are obtained explicitly. The technique resolves the shortcomings of the current mathematical machinery for the calculation of the propagator. For the evaluation of the first two displacement cumulants, the leading order Gaussian propagator is sufficient. However, some functionals of the propagator, such as the entropy production, which we refer to as "first derivatives of the trajectory", need to be evaluated to a previously-unrecognized higher order. The method presented here can be extended to arbitrarily higher orders in order to accurately compute any other functional of the propagator., Comment: 26 pages

In the rapidly progressing field of generative models, the development of efficient and high-fidelity text-to-image diffusion systems represents a significant frontier. This study introduces YaART, a novel production-grade text-to-image cascaded diffusion model aligned to human preferences using Reinforcement Learning from Human Feedback (RLHF). During the development of YaART, we especially focus on the choices of the model and training dataset sizes, the aspects that were not systematically investigated for text-to-image cascaded diffusion models before. In particular, we comprehensively analyze how these choices affect both the efficiency of the training process and the quality of the generated images, which are highly important in practice. Furthermore, we demonstrate that models trained on smaller datasets of higher-quality images can successfully compete with those trained on larger datasets, establishing a more efficient scenario of diffusion models training. From the quality perspective, YaART is consistently preferred by users over many existing state-of-the-art models., Comment: Prompts and additional information are available on the project page, see https://ya.ru/ai/art/paper-yaart-v1

The evolution of random undirected graphs by the clustering attachment (CA) both without node and edge deletion and with uniform node or edge deletion is investigated. Theoretical results are obtained for the CA without node and edge deletion when a newly appended node is connected to two existing nodes of the graph at each evolution step. Theoretical results concern to (1) the sequence of increments of the consecutive mean clustering coefficients tends to zero; (2) the sequences of node degrees and triangle counts of any fixed node which are proved to be submartingales. These results were obtained for any initial graph. The simulation study is provided for the CA with uniform node or edge deletion and without any deletion. It is shown that (1) the CA leads to light-tailed distributed node degrees and triangle counts; (2) the average clustering coefficient tends to a constant over time; (3) the mean node degree and the mean triangle count increase over time with the rate depending on the parameters of the CA. The exposition is accompanied by a real data study., Comment: 25 pages, 9 figures

The use of higher-dimensional photonic encodings (qudits) instead of two-dimensional encodings (qubits) can improve the loss tolerance and reduce the computational resources of photonic-based quantum information processing. To harness this potential, efficient schemes for entangling operations such as the high-dimensional generalization of a linear optics Bell measurement will be required. We show how an efficient high-dimensional entangled state analyzer can be implemented with a linear optics interferometer and auxiliary photonic states. The degree of entanglement of the auxiliary state is much less than in previous protocols as quantified by an exponentially smaller Schmidt rank. In addition, the auxiliary state only occupies a single spatial mode, allowing it to be generated deterministically from a single quantum emitter coupled to a small qubit register. The reduced complexity of the auxiliary states results in a high robustness to imperfections and we show that auxiliary states with fidelities > 0.9 for qudit dimensions 4 can be generated in the presence of qubit error rates on the order of 10%. This paves the way for experimental demonstrations with current hardware., Comment: 13 pages, 5 figures

This paper delves into the significance of the tomographic probability density function (pdf) representation of quantum states, shedding light on the special classes of pdfs that can be tomograms. Instead of using wave functions or density operators on Hilbert spaces, tomograms, which are the true pdfs, are used to completely describe the states of quantum systems. Unlike quasi-pdfs, like the Wigner function, tomograms can be analysed using all the tools of classical probability theory for pdf estimation, which can allow a better quality of state reconstruction. This is particularly useful when dealing with non-Gaussian states where the pdfs are multi-mode. The knowledge of the family of distributions plays an important role in the application of both parametric and non-parametric density estimation methods. We show that not all pdfs can play the role of tomograms of quantum states and introduce the conditions that must be fulfilled by pdfs to be "quantum"., Comment: 21 pages, 3 figures

The clustering attachment model introduced in the paper Bagrow and Brockmann (2013) may be used as an evolution tool of random networks. We propose a new clustering attachment model which can be considered as the limit of the former clustering attachment model as model parameter $\alpha$ tends to zero. We focus on the study of a total triangle count that is considered in the literature as an important characteristic of the network clustering. It is proved that total triangle count tends to infinity a.s. for the proposed model. Our simulation study is used for the modeling of sequences of triangle counts. It is based on the interpretation of the clustering attachment as a generalized P\'{o}lya-Eggenberger urn model that is introduced here at first time., Comment: 16 pages, 2 figures

Energetic and optical properties of plasmonic nanocrystals strongly depend on their sizes, shapes, and composition. Whereas using plasmonic nanoparticles in biotesting has become routine, applications of plasmonics in energy are still early in development. Here, we investigate hot electron (HE) generation and related electromagnetic effects in both mono- and bi-metallic nanorods (NRs) and focus on one promising type of bi-metallic nanocrystals - core-shell Au-Ag nanorods. The spectra of the NRs are broadband, highly tunable with their geometry, and have few plasmon resonances. In this work, we provide a new quantum formalism describing the HE generation in bi-metallic nanostructures. Interestingly, we observe that the HE generation rate at the UV plasmon resonance of Au-Ag NRs appears to be very high. These HEs are highly energetic and suitable for carbon-fuel reactions. Simultaneously, the HE generation at the longitudinal plasmon (L-plasmon) peaks, which can be tuned from the yellow to near-IR, depends on the near-field and electromagnetic Mie effects, limiting the HE efficiencies for the long and large NRs. These properties of the L-plasmon relate to all kinds of NRs (Au, Ag, and Au-Ag). We also consider the generation of the interband d-holes in Au and Ag, since the involvement of the d-band is crucial for the energetic properties of UV plasmons. The proposed formalism is an important development for the description of bi-metallic (or tri-metallic, or more complex) nanostructures, and it paves the way to the efficient application of the plasmonic HEs and hot holes in sensing, nanotechnology, photocatalysis, and electrophotochemistry.

Generating entanglement between distributed network nodes is a prerequisite for the quantum internet. Entanglement distribution protocols based on high-dimensional photonic qudits enable the simultaneous generation of multiple entangled pairs, which can significantly reduce the required coherence time of the qubit registers. However, current schemes require fast optical switching, which is experimentally challenging. In addition, the higher degree of error correlation between the generated entangled pairs in qudit protocols compared to qubit protocols has not been studied in detail. We propose a qudit-mediated entangling protocol that completely circumvents the need for optical switches, making it more accessible for current experimental systems. Furthermore, we quantify the amount of error correlation between the simultaneously generated entangled pairs and analyze the effect on entanglement purification algorithms and teleportation-based quantum error correction. We find that optimized purification schemes can efficiently correct the correlated errors, while the quantum error correction codes studied here perform worse than for uncorrelated error models.

Strain-controlled criticality governs the elasticity of jamming and fiber networks. While the upper critical dimension of jamming is believed to be $d_u$=2, non mean-field exponents are observed in numerical studies of 2D and 3D fiber networks. The origins of this remains unclear. In this study we propose a minimal mean-field model for strain-controlled criticality of fiber networks. We then extend this to a phenomenological field theory, in which non mean-field behavior emerges as a result of the disorder in the network structure. We predict that the upper critical dimension for such systems is $d_u$=4 using a Gaussian approximation. Moreover, we identify an order parameter for the phase transition, which has been lacking for fiber networks to date.

We present the first in-depth empirical characterization of the costs of trading on a decentralized exchange (DEX). Using quoted prices from the Uniswap Labs interface for two pools -- USDC-ETH (5bps) and PEPE-ETH (30bps) -- we evaluate the efficiency of trading on DEXs. Our main tool is slippage -- the difference between the realized execution price of a trade, and its quoted price -- which we breakdown into its benign and adversarial components. We also present an alternative way to quantify and identify slippage due to adversarial reordering of transactions, which we call reordering slippage, that does not require quoted prices or mempool data to calculate. We find that the composition of transaction costs varies tremendously with the trade's characteristics. Specifically, while for small swaps, gas costs dominate costs, for large swaps price-impact and slippage account for the majority of it. Moreover, when trading PEPE, a popular 'memecoin', the probability of adversarial slippage is about 80% higher than when trading a mature asset like USDC. Overall, our results provide preliminary evidence that DEXs offer a compelling trust-less alternative to centralized exchanges for trading digital assets., Comment: 31 pages, 7 tables, 2 figures

The phase shift rules enable the estimation of the derivative of a quantum state with respect to phase parameters, providing valuable insights into the behavior and dynamics of quantum systems. This capability is essential in quantum simulation tasks where understanding the behavior of complex quantum systems is of interest, such as simulating chemical reactions or condensed matter systems. However, parameter shift rules are typically designed for Hamiltonian systems with equidistant eigenvalues. For systems with closely spaced eigenvalues, effective rules have not been established. We provide insights about the optimal design of a parameter shift rule, tailored to various sorts of spectral information that may be available. The proposed method lets derivatives be calculated for any system, regardless of how close the eigenvalues are to each other. It also optimizes the number of phase shifts, which reduces the amount of gate resources needed., Comment: 24 pages, 2 figures

The estimation of multi-qubit observables is a key task in quantum information science. The standard approach is to decompose a multi-qubit observable into a weighted sum of Pauli strings. The observable can then be estimated from projective single qubit measurements according to the Pauli strings followed by a classical summation. As the number of Pauli strings in the decomposition increases, shot-noise drastically builds up, and the accuracy of such estimation can be considerably compromised. Access to a single qubit quantum memory, where measurement data may be stored and accumulated can circumvent the build-up of shot noise. Here, we describe a many-qubit observable estimation approach to achieve this with a much lower number of interactions between the multi-qubit device and the single qubit memory compared to previous approaches. Our algorithm offers a reduction in the required number of measurements for a given target variance that scales $N^{\frac{2}{3}}$ with the number of Pauli strings $N$ in the observable decomposition. The low number of interactions between the multi-qubit device and the memory is desirable for noisy intermediate-scale quantum devices., Comment: 20 pages, 4 figures, 1 table. arXiv admin note: text overlap with arXiv:2212.07710

Graph convolutional neural networks (GCNs) operate by aggregating messages over local neighborhoods given the prediction task under interest. Many GCNs can be understood as a form of generalized diffusion of input features on the graph, and significant work has been dedicated to improving predictive accuracy by altering the ways of message passing. In this work, we propose a new convolution kernel that effectively rewires the graph according to the occupation correlations of the vertices by trading on the generalized diffusion paradigm for the propagation of a quantum particle over the graph. We term this new convolution kernel the Quantum Diffusion Convolution (QDC) operator. In addition, we introduce a multiscale variant that combines messages from the QDC operator and the traditional combinatorial Laplacian. To understand our method, we explore the spectral dependence of homophily and the importance of quantum dynamics in the construction of a bandpass filter. Through these studies, as well as experiments on a range of datasets, we observe that QDC improves predictive performance on the widely used benchmark datasets when compared to similar methods.

Networks of stiff fibers govern the elasticity of biological structures such as the extracellular matrix of collagen. These networks are known to stiffen nonlinearly under shear or extensional strain. Recently, it has been shown that such stiffening is governed by a strain-controlled athermal but critical phase transition, from a floppy phase below the critical strain to a rigid phase above the critical strain. While this phase transition has been extensively studied numerically and experimentally, a complete analytical theory for this transition remains elusive. Here, we present an effective medium theory (EMT) for this mechanical phase transition of fiber networks. We extend a previous EMT appropriate for linear elasticity to incorporate nonlinear effects via an anharmonic Hamiltonian. The mean-field predictions of this theory, including the critical exponents, scaling relations and non-affine fluctuations qualitatively agree with previous experimental and numerical results.

Chirality, either of light or matter, has proved to be very practical in biosensing and nanophotonics. However, the fundamental understanding of its temporal dynamics still needs to be discovered. A realistic setup for this are the so-called metastructures, since they are optically active and are built massively, hence rendering an immediate potential candidate. Here we propose and study the electromagnetic-optical mechanism leading to chiral optical imprinting on metastructures. Induced photothermal responses create anisotropic permittivity modulations, different for left or right circularly polarized light, leading to temporal-dependent chiral imprinting of hot-spots, namely imprinting of chirality. The above effect has not been observed yet, but it is within reach of modern experimental approaches. The proposed nonlinear chiroptical effect is general and should appear in any anisotropic material; however, we need to design a particular geometry for this effect to be strong. These new chiral time-dependent metastructures may lead to a plethora of applications., Comment: Main (29 pages, 6 figures) and supplemental (46 pages, 35 figures)

Recommendation systems are a core feature of social media companies with their uses including recommending organic and promoted contents. Many modern recommendation systems are split into multiple stages - candidate generation and heavy ranking - to balance computational cost against recommendation quality. We focus on the candidate generation phase of a large-scale ads recommendation problem in this paper, and present a machine learning first heterogeneous re-architecture of this stage which we term TwERC. We show that a system that combines a real-time light ranker with sourcing strategies capable of capturing additional information provides validated gains. We present two strategies. The first strategy uses a notion of similarity in the interaction graph, while the second strategy caches previous scores from the ranking stage. The graph based strategy achieves a 4.08% revenue gain and the rankscore based strategy achieves a 1.38% gain. These two strategies have biases that complement both the light ranker and one another. Finally, we describe a set of metrics that we believe are valuable as a means of understanding the complex product trade offs inherent in industrial candidate generation systems., Comment: 10 pages, 3 figures

We introduce the concept of Almost-Companion Matrix (ACM) by relaxing the non-derogatory property of the standard Companion Matrix (CM). That is, we define an ACM as a matrix whose characteristic polynomial coincides with a given monic and generally complex polynomial. The greater flexibility inherent in the ACM concept, compared to CM, allows the construction of ACMs that have convenient matrix structures satisfying desired additional conditions, compatibly with specific properties of the polynomial coefficients. We demonstrate the construction of Hermitian and Unitary ACMs starting from appropriate third degree polynomials, with implications for their use in physical-mathematical problems such as the parameterization of the Hamiltonian, density, or evolution matrix of a qutrit. We show that the ACM provides a means of identifying properties of a given polynomial and finding its roots. For example, we describe the ACM-based solution of cubic complex algebraic equations without resorting to the use of the Cardano-Dal Ferro formulas. We also show the necessary and sufficient conditions on the coefficients of a polynomial for it to represent the characteristic polynomial of a unitary ACM. The presented approach can be generalized to complex polynomials of higher degree., Comment: 26 pages

The estimation of many-qubit observables is an essential task of quantum information processing. The generally applicable approach is to decompose the observables into weighted sums of multi-qubit Pauli strings, i.e., tensor products of single-qubit Pauli matrices, which can readily be measured with single qubit rotations. The accumulation of shot noise in this approach, however, severely limits the achievable variance for a finite number of measurements. We introduce a novel method, dubbed Coherent Pauli Summation (CPS) that circumvents this limitation by exploiting access to a single-qubit quantum memory in which measurement information can be stored and accumulated. Our algorithm offers a reduction in the required number of measurements for a given variance that scales linearly with the number of Pauli strings of the decomposed observable. Our work demonstrates how a single long-coherence qubit memory can assist the operation of noisy many-qubit quantum devices in a cardinal task., Comment: 2 figures

A network evolution with predicted tail and extremal indices of PageRank and the Max-Linear Model used as node influence indices in random graphs is considered. The tail index shows a heaviness of the distribution tail. The extremal index is a measure of clustering (or local dependence) of the stochastic process. The cluster implies a set of consecutive exceedances of the process over a sufficiently high threshold. Our recent results concerning sums and maxima of non-stationary random length sequences of regularly varying random variables are extended to random graphs. Starting with a set of connected stationary seed communities as a hot spot and ranking them with regard to their tail indices, the tail and extremal indices of new nodes that are appended to the network may be determined. This procedure allows us to predict a temporal network evolution in terms of tail and extremal indices. The extremal index determines limiting distributions of a maximum of the PageRank and the Max-Linear Model of newly attached nodes. The exposition is provided by algorithms and examples. To validate our theoretical results, our simulation and real data study concerning a linear preferential attachment as a tool for network growth are provided.

Odd viscosity couples stress to strain rate in a dissipationless way. It has been studied in plasmas under magnetic fields, superfluid ${\rm He}^3$, quantum-Hall fluids, and recently in the context of chiral active matter. In most of these studies, odd terms in the viscosity obey Onsager reciprocal relations. Although this is expected in equilibrium systems, it is not obvious that Onsager relations hold in active materials. By directly coarse graining the kinetic energy and independently using both the Poisson-bracket formalism and a kinetic theory derivation, we find that the appearance of a non-vanishing angular momentum density, which is a hallmark of chiral active materials, necessarily breaks Onsager reciprocal relations. This leads to a non-Hermitian dynamical matrix for the total hydrodynamic momentum and to the appearance of odd viscosity and other non-dissipative contributions to the viscosity. Furthermore, by accounting for both the angular momentum density and interactions that lead to odd viscosity, we find regions in the parameter space in which 3D odd mechanical waves propagate and regions in which they are mechanically unstable. The lines separating these regions are continuous lines of exceptional points, suggesting a possible non-reciprocal phase transition., Comment: 12 pages, 4 figures, SI

Graph attention networks estimate the relational importance of node neighbors to aggregate relevant information over local neighborhoods for a prediction task. However, the inferred attentions are vulnerable to spurious correlations and connectivity in the training data, hampering the generalizability of the model. We introduce CAR, a general-purpose regularization framework for graph attention networks. Embodying a causal inference approach, CAR aligns the attention mechanism with the causal effects of active interventions on graph connectivity in a scalable manner. CAR is compatible with a variety of graph attention architectures, and we show that it systematically improves generalizability on various node classification tasks. Our ablation studies indicate that CAR hones in on the aspects of graph structure most pertinent to the prediction (e.g., homophily), and does so more effectively than alternative approaches. Finally, we also show that CAR enhances interpretability of attention weights by accentuating node-neighbor relations that point to causal hypotheses. For social media network-sized graphs, a CAR-guided graph rewiring approach could allow us to combine the scalability of graph convolutional methods with the higher performance of graph attention.

Many Graph Neural Networks (GNNs) perform poorly compared to simple heuristics on Link Prediction (LP) tasks. This is due to limitations in expressive power such as the inability to count triangles (the backbone of most LP heuristics) and because they can not distinguish automorphic nodes (those having identical structural roles). Both expressiveness issues can be alleviated by learning link (rather than node) representations and incorporating structural features such as triangle counts. Since explicit link representations are often prohibitively expensive, recent works resorted to subgraph-based methods, which have achieved state-of-the-art performance for LP, but suffer from poor efficiency due to high levels of redundancy between subgraphs. We analyze the components of subgraph GNN (SGNN) methods for link prediction. Based on our analysis, we propose a novel full-graph GNN called ELPH (Efficient Link Prediction with Hashing) that passes subgraph sketches as messages to approximate the key components of SGNNs without explicit subgraph construction. ELPH is provably more expressive than Message Passing GNNs (MPNNs). It outperforms existing SGNN models on many standard LP benchmarks while being orders of magnitude faster. However, it shares the common GNN limitation that it is only efficient when the dataset fits in GPU memory. Accordingly, we develop a highly scalable model, called BUDDY, which uses feature precomputation to circumvent this limitation without sacrificing predictive performance. Our experiments show that BUDDY also outperforms SGNNs on standard LP benchmarks while being highly scalable and faster than ELPH., Comment: 29 pages, 19 figures, 6 appendices

The sums and maxima of weighted non-stationary random length sequences of regularly varying random variables may have the same tail and extremal indices, Markovich and Rodionov (2020). The main constraints are that there exists a unique series in a scheme of series with the minimum tail index, the tail of the term number is lighter than the tail of the terms and the weights are positive constants. These assumptions are changed here: a bounded random number of series is allowed to have the minimum tail index, the tail of the term number may be heavier than the tail of the terms and the weights may be real-valued. Then we derive the tail and extremal indices of the weighted non-stationary random length sequences under the new assumptions., Comment: 23 pages