Search

The Kirkwood-Dirac quasiprobability distribution emerges from the quantum correlation function of two observables measured at distinct times and is therefore relevant for fundamental physics and quantum technologies. These quasiprobabilities follow all but one of Kolmogorov axioms for joint probability distributions: they can take non-positive values. Their experimental reconstruction becomes challenging when expectation values of incompatible observables are involved. Previous strategies aimed to reconstruct them using weak measurements or combining strong measurements. Here, we use a more direct approach, an interferometric scheme aided by an auxiliary system, to reconstruct the Kirkwood-Dirac quasiprobability distribution. We experimentally demonstrate the interferometric scheme in an electron-nuclear spin system associated with a nitrogen-vacancy center in diamond. By measuring the characteristic function, we reconstruct the quasiprobability distribution of the work and analyze the behavior of the first and second moments of work. Our results clarify the physical meaning of the work quasiprobability distribution in the context of quantum thermodynamics. Finally, having measured the real and imaginary parts of the Kirkwood-Dirac quasiprobability of work, we are also able to study the uncertainty of measuring the Hamiltonian of the system at two times, via the Robertson-Schr{\"o}dinger uncertainty relation, for different initial states., Comment: 15 pages, 8 figures

In the evolving landscape of deep learning, selecting the best pre-trained models from a growing number of choices is a challenge. Transferability scorers propose alleviating this scenario, but their recent proliferation, ironically, poses the challenge of their own assessment. In this work, we propose both robust benchmark guidelines for transferability scorers, and a well-founded technique to combine multiple scorers, which we show consistently improves their results. We extensively evaluate 13 scorers from literature across 11 datasets, comprising generalist, fine-grained, and medical imaging datasets. We show that few scorers match the predictive performance of the simple raw metric of models on ImageNet, and that all predictors suffer on medical datasets. Our results highlight the potential of combining different information sources for reliably predicting transferability across varied domains., Comment: 15 pages, 3 figures, 2 tables

Growing interest and investment in the capabilities of foundation models has positioned such systems to impact a wide array of public services. Alongside these opportunities is the risk that these systems reify existing power imbalances and cause disproportionate harm to marginalized communities. Participatory approaches hold promise to instead lend agency and decision-making power to marginalized stakeholders. But existing approaches in participatory AI/ML are typically deeply grounded in context - how do we apply these approaches to foundation models, which are, by design, disconnected from context? Our paper interrogates this question. First, we examine existing attempts at incorporating participation into foundation models. We highlight the tension between participation and scale, demonstrating that it is intractable for impacted communities to meaningfully shape a foundation model that is intended to be universally applicable. In response, we develop a blueprint for participatory foundation models that identifies more local, application-oriented opportunities for meaningful participation. In addition to the "foundation" layer, our framework proposes the "subfloor'' layer, in which stakeholders develop shared technical infrastructure, norms and governance for a grounded domain, and the "surface'' layer, in which affected communities shape the use of a foundation model for a specific downstream task. The intermediate "subfloor'' layer scopes the range of potential harms to consider, and affords communities more concrete avenues for deliberation and intervention. At the same time, it avoids duplicative effort by scaling input across relevant use cases. Through three case studies in clinical care, financial services, and journalism, we illustrate how this multi-layer model can create more meaningful opportunities for participation than solely intervening at the foundation layer., Comment: 13 pages, 2 figures. Appeared at FAccT '24

The two-dimensional electron gas (2DEG) is a fundamental model, which is drawing increasing interest because of recent advances in experimental and theoretical studies of 2D materials. Current understanding of the ground state of the 2DEG relies on quantum Monte Carlo calculations, based on variational comparisons of different ansatze for different phases. We use a single variational ansatz, a general backflow-type wave function using a message-passing neural quantum state architecture, for a unified description across the entire density range. The variational optimization consistently leads to lower ground-state energies than previous best results. Transition into a Wigner crystal (WC) phase occurs automatically at rs = 37 +/- 1, a density lower than currently believed. Between the liquid and WC phases, the same ansatz and variational search strongly suggest the existence of intermediate states in a broad range of densities, with enhanced short-range nematic spin correlations.

Image classification is one of the most fundamental capabilities of machine vision intelligence. In this work, we revisit the image classification task using visually-grounded language models (VLMs) such as GPT-4V and LLaVA. We find that existing proprietary and public VLMs, despite often using CLIP as a vision encoder and having many more parameters, significantly underperform CLIP on standard image classification benchmarks like ImageNet. To understand the reason, we explore several hypotheses concerning the inference algorithms, training objectives, and data processing in VLMs. Our analysis reveals that the primary cause is data-related: critical information for image classification is encoded in the VLM's latent space but can only be effectively decoded with enough training data. Specifically, there is a strong correlation between the frequency of class exposure during VLM training and instruction-tuning and the VLM's performance in those classes; when trained with sufficient data, VLMs can match the accuracy of state-of-the-art classification models. Based on these findings, we enhance a VLM by integrating classification-focused datasets into its training, and demonstrate that the enhanced classification performance of the VLM transfers to its general capabilities, resulting in an improvement of 11.8% on the newly collected ImageWikiQA dataset., Comment: Published at NeurIPS 2024

The task of Visual Place Recognition (VPR) is to predict the location of a query image from a database of geo-tagged images. Recent studies in VPR have highlighted the significant advantage of employing pre-trained foundation models like DINOv2 for the VPR task. However, these models are often deemed inadequate for VPR without further fine-tuning on task-specific data. In this paper, we propose a simple yet powerful approach to better exploit the potential of a foundation model for VPR. We first demonstrate that features extracted from self-attention layers can serve as a powerful re-ranker for VPR. Utilizing these features in a zero-shot manner, our method surpasses previous zero-shot methods and achieves competitive results compared to supervised methods across multiple datasets. Subsequently, we demonstrate that a single-stage method leveraging internal ViT layers for pooling can generate global features that achieve state-of-the-art results, even when reduced to a dimensionality as low as 128D. Nevertheless, incorporating our local foundation features for re-ranking, expands this gap. Our approach further demonstrates remarkable robustness and generalization, achieving state-of-the-art results, with a significant gap, in challenging scenarios, involving occlusion, day-night variations, and seasonal changes.

Personalized retrieval and segmentation aim to locate specific instances within a dataset based on an input image and a short description of the reference instance. While supervised methods are effective, they require extensive labeled data for training. Recently, self-supervised foundation models have been introduced to these tasks showing comparable results to supervised methods. However, a significant flaw in these models is evident: they struggle to locate a desired instance when other instances within the same class are presented. In this paper, we explore text-to-image diffusion models for these tasks. Specifically, we propose a novel approach called PDM for Personalized Features Diffusion Matching, that leverages intermediate features of pre-trained text-to-image models for personalization tasks without any additional training. PDM demonstrates superior performance on popular retrieval and segmentation benchmarks, outperforming even supervised methods. We also highlight notable shortcomings in current instance and segmentation datasets and propose new benchmarks for these tasks., Comment: Accepted to NeurIPS 2024

We introduce an innovative method for incremental nonparametric probabilistic inference in high-dimensional state spaces. Our approach leverages \slices from high-dimensional surfaces to efficiently approximate posterior distributions of any shape. Unlike many existing graph-based methods, our \slices perspective eliminates the need for additional intermediate reconstructions, maintaining a more accurate representation of posterior distributions. Additionally, we propose a novel heuristic to balance between accuracy and efficiency, enabling real-time operation in nonparametric scenarios. In empirical evaluations on synthetic and real-world datasets, our \slices approach consistently outperforms other state-of-the-art methods. It demonstrates superior accuracy and achieves a significant reduction in computational complexity, often by an order of magnitude., Comment: 8 Pages, 7 figures, Submitted to IEEE IROS 2024

In this paper, we investigate the challenging framework of Byzantine-robust training in distributed machine learning (ML) systems, focusing on enhancing both efficiency and practicality. As distributed ML systems become integral for complex ML tasks, ensuring resilience against Byzantine failures-where workers may contribute incorrect updates due to malice or error-gains paramount importance. Our first contribution is the introduction of the Centered Trimmed Meta Aggregator (CTMA), an efficient meta-aggregator that upgrades baseline aggregators to optimal performance levels, while requiring low computational demands. Additionally, we propose harnessing a recently developed gradient estimation technique based on a double-momentum strategy within the Byzantine context. Our paper highlights its theoretical and practical advantages for Byzantine-robust training, especially in simplifying the tuning process and reducing the reliance on numerous hyperparameters. The effectiveness of this technique is supported by theoretical insights within the stochastic convex optimization (SCO) framework and corroborated by empirical evidence.

The Data Acquisition System (DAQ) for the LUX-ZEPLIN (LZ) dark matter detector is described. The signals from 745 PMTs, distributed across three subsystems, are sampled with 100-MHz 32-channel digitizers (DDC-32s). A basic waveform analysis is carried out on the on-board Field Programmable Gate Arrays (FPGAs) to extract information about the observed scintillation and electroluminescence signals. This information is used to determine if the digitized waveforms should be preserved for offline analysis. The system is designed around the Kintex-7 FPGA. In addition to digitizing the PMT signals and providing basic event selection in real time, the flexibility provided by the use of FPGAs allows us to monitor the performance of the detector and the DAQ in parallel to normal data acquisition. The hardware and software/firmware of this FPGA-based Architecture for Data acquisition and Realtime monitoring (FADR) are discussed and performance measurements are described., Comment: 18 pages, 24 figures

This article introduces a general mesh intersection algorithm that exactly computes the so-called Weiler model and that uses it to implement boolean operations with arbitrary multi-operand expressions, CSG (constructive solid geometry) and some mesh repair operations. From an input polygon soup, the algorithm first computes the co-refinement, with an exact representation of the intersection points. Then, the decomposition of 3D space into volumetric regions (Weiler model) is constructed, by sorting the facets around the non-manifold intersection edges (radial sort), using specialized exact predicates. Finally, based on the input boolean expression, the triangular facets that belong to the boundary of the result are classified. This is, to our knowledge, the first algorithm that computes an exact Weiler model. To implement all the involved predicates and constructions, two geometric kernels are proposed, tested and discussed (arithmetic expansions and multi-precision floating-point). As a guiding principle,the combinatorial information shared between each step is kept as simple as possible. It is made possible by treating all the particular cases in the kernel. In particular, triangles with intersections are remeshed using the (uniquely defined) Constrained Delaunay Triangulation, with symbolic perturbations to disambiguate configurations with co-cyclic points. It makes it easy to discard the duplicated triangles that appear when remeshing overlapping facets. The method is tested and compared with previous work, on the existing "thingi10K" dataset (to test co-refinement and mesh repair) and on a new "thingiCSG" dataset made publicly available (to test the full CSG pipeline) on a variety of interesting examples featuring different types of "pathologies", Comment: 25 pages, 20 figures

The SALTUS Probe mission will provide a powerful far-infrared (far-IR) pointed space observatory to explore our cosmic origins and the possibility of life elsewhere. The observatory employs an innovative deployable 14-m aperture, with a sunshield that will radiatively cool the off-axis primary to <45K. This cooled primary reflector works in tandem with cryogenic coherent and incoherent instruments that span the 34 to 660 micron far-IR range at both high and moderate spectral resolutions., Comment: 49 pages, 10 figures, 3 tables, submitted to SPIE JATIS

The ability to build and leverage world models is essential for a general-purpose AI agent. Testing such capabilities is hard, in part because the building blocks of world models are ill-defined. We present Elements of World Knowledge (EWOK), a framework for evaluating world modeling in language models by testing their ability to use knowledge of a concept to match a target text with a plausible/implausible context. EWOK targets specific concepts from multiple knowledge domains known to be vital for world modeling in humans. Domains range from social interactions (help/hinder) to spatial relations (left/right). Both, contexts and targets are minimal pairs. Objects, agents, and locations in the items can be flexibly filled in enabling easy generation of multiple controlled datasets. We then introduce EWOK-CORE-1.0, a dataset of 4,374 items covering 11 world knowledge domains. We evaluate 20 openweights large language models (1.3B--70B parameters) across a battery of evaluation paradigms along with a human norming study comprising 12,480 measurements. The overall performance of all tested models is worse than human performance, with results varying drastically across domains. These data highlight simple cases where even large models fail and present rich avenues for targeted research on LLM world modeling capabilities., Comment: 21 pages (11 main), 7 figures. Authors Anna Ivanova, Aalok Sathe, Benjamin Lipkin contributed equally

One hallmark of human language is its combinatoriality -- reusing a relatively small inventory of building blocks to create a far larger inventory of increasingly complex structures. In this paper, we explore the idea that combinatoriality in language reflects a human inductive bias toward representational efficiency in symbol systems. We develop a computational framework for discovering structure in a writing system. Built on top of state-of-the-art library learning and program synthesis techniques, our computational framework discovers known linguistic structures in the Chinese writing system and reveals how the system evolves towards simplification under pressures for representational efficiency. We demonstrate how a library learning approach, utilizing learned abstractions and compression, may help reveal the fundamental computational principles that underlie the creation of combinatorial structures in human cognition, and offer broader insights into the evolution of efficient communication systems., Comment: Accepted at CogSci 2024 (Talk)

Given a natural number $n$, let $\omega\left(n\right)$ denote the number of distinct prime factors of $n$, let $Z$ denote a standard normal variable, and let $P_{n}$ denote the uniform distribution on $\left\{ 1,\ldots,n\right\} $. The Erd\H{o}s-Kac Theorem states that if $N\left(n\right)$ is a uniformly distributed variable on $\lbrace 1,\ldots,n \rbrace$, then $\omega\left(N\left(n\right)\right)$ is asymptotically normally distributed as $n\to \infty$ with both mean and variance equal to $\log \log n$. The contribution of this paper is a generalization of the Erd\H{o}s-Kac Theorem to a larger class of random variables by considering perturbations of the uniform probability mass $1/n$ in the following sense. Denote by $\mathbb{P}_{n}$ a probability distribution on $\left\{ 1,\ldots,n\right\} $ given by $\mathbb{P}_{n}\left(i\right)=1/n+\varepsilon_{i,n}$. We provide sufficient conditions on $\varepsilon_{i,n}$ so that the number of distinct prime factors of a $\mathbb{P}_{n}$-distributed random variable is asymptotically normally distributed, as $n\to \infty$, with both mean and variance equal to $\log \log n$. Our main result is applied to prove that the number of distinct prime factors of a positive integer with the Harmonic$\left(n\right)$ distribution also tends to the normal distribution, as $n\to \infty$. In addition, we explore sequences of distributions on the natural numbers such that $\omega(n)$ is normally distributed in the limit. In addition, one of our theorems and its corollaries generalize a result from the literature involving the limit of $Zeta\left(s\right)$ distributions as the parameter $s \to 1$., Comment: Supersedes arXiv:2011.00152v1

We propose an interactive approach to language learning that utilizes linguistic acceptability judgments from an informant (a competent language user) to learn a grammar. Given a grammar formalism and a framework for synthesizing data, our model iteratively selects or synthesizes a data-point according to one of a range of information-theoretic policies, asks the informant for a binary judgment, and updates its own parameters in preparation for the next query. We demonstrate the effectiveness of our model in the domain of phonotactics, the rules governing what kinds of sound-sequences are acceptable in a language, and carry out two experiments, one with typologically-natural linguistic data and another with a range of procedurally-generated languages. We find that the information-theoretic policies that our model uses to select items to query the informant achieve sample efficiency comparable to, and sometimes greater than, fully supervised approaches.

Millisecond pulsars with white dwarf companions have typical eccentricities $e\sim 10^{-6}-10^{-3}$. The eccentricities of helium white dwarfs are explained well by applying the fluctuation-dissipation theorem to convective eddies in their red giant progenitors. We extend this theory to more massive carbon-oxygen (CO) white dwarfs with asymptotic giant branch (AGB) progenitors. Due to the radiation pressure in AGB stars, the dominant factor in determining the remnant white dwarf's eccentricity is the critical residual hydrogen envelope mass $m_{\rm env}$ required to inflate the star to giant proportions. Using a suite of MESA stellar evolution simulations with $\Delta m_{\rm c}=10^{-3}\,{\rm M}_\odot$ core-mass intervals, we resolved the AGB thermal pulses and found that the critical $m_{\rm env}\propto m_{\rm c}^{-6}$. The resulting eccentricity $e\sim 3\times 10^{-3}$ is almost independent of the remnant CO white dwarf's mass $m_{\rm c}$. Nearly all of the measured eccentricities lie below this robust theoretical limit, indicating that the eccentricity is damped during the common-envelope inspiral that follows the unstable Roche-lobe overflow of the AGB star. Specifically, we focused on white dwarfs with median masses $m_{\rm c}>0.6\,{\rm M}_\odot$. These massive white dwarfs begin their inspiral with practically identical orbital periods and eccentricities, eliminating any dependence on the initial conditions. For this sub-sample, we find an empirical relation $e\propto P^{3/2}$ between the final period and eccentricity that is much tighter than previous studies - motivating theoretical work on the eccentricity evolution during the common envelope phase. The eccentricities of lower mass CO white dwarfs may be explained by alternative formation channels., Comment: to appear in MNRAS, new Section 5 and Appendix

In this paper we study the filamentary substructure of 3.3 $\mu$m PAH emission from JWST/NIRCam observations in the base of the M82 star-burst driven wind. We identify plume-like substructure within the PAH emission with widths of $\sim$50 pc. Several of the plumes extend to the edge of the field-of-view, and thus are at least 200-300 pc in length. In this region of the outflow, the vast majority ($\sim$70\%) of PAH emission is associated with the plumes. We show that those structures contain smaller scale "clouds" with widths that are $\sim$5-15 pc, and they are morphologically similar to the results of "cloud-crushing" simulations. We estimate the cloud-crushing time-scales of $\sim$0.5-3 Myr, depending on assumptions. We show this time scale is consistent with a picture in which these observed PAH clouds survived break-out from the disk rather than being destroyed by the hot wind. The PAH emission in both the midplane and the outflow is shown to tightly correlate with that of Pa$\alpha$ emission (from HST/NICMOS data), at the scale of both plumes and clouds, though the ratio of PAH-to-Pa$\alpha$ increases at further distances from the midplane. Finally, we show that the outflow PAH emission is suppressed in regions of the M82 wind that are bright in X-ray emission. Overall, our results are broadly consistent with a picture in which cold gas in galactic outflows is launched via hierarchically structured plumes, and those small scale clouds are more likely to survive the wind environment when collected into the larger plume structure., Comment: Submitted to MNRAS

LUX-ZEPLIN (LZ) is a tonne-scale experiment searching for direct dark matter interactions and other rare events. It is located at the Sanford Underground Research Facility (SURF) in Lead, South Dakota, USA. The core of the LZ detector is a dual-phase xenon time projection chamber (TPC), designed with the primary goal of detecting Weakly Interacting Massive Particles (WIMPs) via their induced low energy nuclear recoils. Surrounding the TPC, two veto detectors immersed in an ultra-pure water tank enable reducing background events to enhance the discovery potential. Intricate calibration systems are purposely designed to precisely understand the responses of these three detector volumes to various types of particle interactions and to demonstrate LZ's ability to discriminate between signals and backgrounds. In this paper, we present a comprehensive discussion of the key features, requirements, and performance of the LZ calibration systems, which play a crucial role in enabling LZ's WIMP-search and its broad science program. The thorough description of these calibration systems, with an emphasis on their novel aspects, is valuable for future calibration efforts in direct dark matter and other rare-event search experiments.

Vision-based end-to-end driving models trained by imitation learning can lead to affordable solutions for autonomous driving. However, training these well-performing models usually requires a huge amount of data, while still lacking explicit and intuitive activation maps to reveal the inner workings of these models while driving. In this paper, we study how to guide the attention of these models to improve their driving quality and obtain more intuitive activation maps by adding a loss term during training using salient semantic maps. In contrast to previous work, our method does not require these salient semantic maps to be available during testing time, as well as removing the need to modify the model's architecture to which it is applied. We perform tests using perfect and noisy salient semantic maps with encouraging results in both, the latter of which is inspired by possible errors encountered with real data. Using CIL++ as a representative state-of-the-art model and the CARLA simulator with its standard benchmarks, we conduct experiments that show the effectiveness of our method in training better autonomous driving models, especially when data and computational resources are scarce., Comment: Accepted for publication at the 35th IEEE Intelligent Vehicles Symposium (IV 2024)

In this article, we prove that Ramsey's theorem for pairs and two colors is a $\forall \Pi^0_4$ conservative extension of $\mathsf{RCA}_0 + \mathsf{B}\Sigma^0_2$, where a $\forall \Pi^0_4$ formula consists of a universal quantifier over sets followed by a $\Pi^0_4$ formula. The proof is an improvement of a result by Patey and Yokoyama and a step towards the resolution of the longstanding question of the first-order part of Ramsey's theorem for pairs., Comment: 36 pages. Changed the definition of Ramsey-like-Pi12-statement to the more restrictive notion of RT-like statement

A left-variable word over an alphabet~$A$ is a word over~$A \cup \{\star\}$ whose first letter is the distinguished symbol~$\star$ standing for a placeholder. The Ordered Variable Word theorem ($\mathsf{OVW}$), also known as Carlson-Simpson's theorem, is a tree partition theorem, stating that for every finite alphabet~$A$ and every finite coloring of the words over~$A$, there exists a word $c_0$ and an infinite sequence of left-variable words $w_1, w_2, \dots$ such that $\{ c_0 \cdot w_1[a_1] \cdot \dots \cdot w_k[a_k] : k \in \mathbb{N}, a_1, \dots, a_k \in A \}$ is monochromatic. In this article, we prove that $\mathsf{OVW}$ is $\Pi^0_4$-conservative over~$\mathsf{RCA}_0 + \mathsf{B}\Sigma^0_2$. This implies in particular that $\mathsf{OVW}$ does not imply $\mathsf{ACA}_0$ over~$\mathsf{RCA}_0$. This is the first principle for which the only known separation from~$\mathsf{ACA}_0$ involves non-standard models., Comment: 19 pages

The first science run of the LUX-ZEPLIN (LZ) experiment, a dual-phase xenon time project chamber operating in the Sanford Underground Research Facility in South Dakota, USA, has reported leading limits on spin-independent WIMP-nucleon interactions and interactions described from a non-relativistic effective field theory (NREFT). Using the same 5.5~t fiducial mass and 60 live days of exposure we report on the results of a relativistic extension to the NREFT. We present constraints on couplings from covariant interactions arising from the coupling of vector, axial currents, and electric dipole moments of the nucleon to the magnetic and electric dipole moments of the WIMP which cannot be described by recasting previous results described by an NREFT. Using a profile-likelihood ratio analysis, in an energy region between 0~keV$_\text{nr}$ to 270~keV$_\text{nr}$, we report 90% confidence level exclusion limits on the coupling strength of five interactions in both the isoscalar and isovector bases., Comment: 7 pages, 4 figures

This paper presents an overview of the Milky Way and nearby galaxies science case for the \textit{Single Aperture Large Telescope for Universe Studies} (SALTUS) far-infrared NASA probe-class mission concept. SALTUS offers enormous gains in spatial resolution and spectral sensitivity over previous far-IR missions, thanks to its cold ($<$40~K) 14-m primary mirror. Key Milky Way and nearby galaxies science goals for SALTUS focus on understanding the role of star formation in feedback in the Local Universe. In addition to this science case, SALTUS would open a new window to to of Galactic and extragalactic communities in the 2030s, enable fundamentally new questions to be answered, and be a far-IR analog to the near- and mid-IR capabilities of JWST. This paper summarizes the Milky Way and nearby galaxies science case and plans for notional observing programs in both guaranteed and guest (open) time., Comment: Resubmitted to JATIS as part of a special issue on SALTUS

This paper presents an overview of the high-redshift extragalactic science case for the Single Aperture Large Telescope for Universe Studies (SALTUS) far-infrared NASA probe-class mission concept. Enabled by its 14m primary reflector, SALTUS offers enormous gains in spatial resolution and spectral sensitivity over previous far-IR missions. SALTUS would be a versatile observatory capable of responding to the scientific needs of the extragalactic community in the 2030s, and a natural follow-on to the near- and mid-IR capabilities of JWST. Key early-universe science goals for SALTUS focus on understanding the role of galactic feedback processes in regulating galaxy growth across cosmic time, and charting the rise of metals and dust from the early universe to the present. This paper summarizes these science cases and the performance metrics most relevant for high-redshift observations., Comment: Submitted to JATIS as part of a special issue on SALTUS. Key performance metrics in Figs. 1 and 2

Label-free imaging through two-photon autofluorescence (2PAF) of NAD(P)H allows for non-destructive and high-resolution visualization of cellular activities in living systems. However, its application to thick tissues and organoids has been restricted by its limited penetration depth within 300 $\mu$m, largely due to tissue scattering at the typical excitation wavelength (~750 nm) required for NAD(P)H. Here, we demonstrate that the imaging depth for NAD(P)H can be extended to over 700 $\mu$m in living engineered human multicellular microtissues by adopting multimode fiber (MMF)-based low-repetition-rate high-peak-power three-photon (3P) excitation of NAD(P)H at 1100 nm. This is achieved by having over 0.5 MW peak power at the band of 1100$\pm$25 nm through adaptively modulating multimodal nonlinear pulse propagation with a compact fiber shaper. Moreover, the 8-fold increase in pulse energy at 1100 nm enables faster imaging of monocyte behaviors in the living multicellular models. These results represent a significant advance for deep and dynamic metabolic and structural imaging of intact living biosystems. The modular design (MMF with a slip-on fiber shaper) is anticipated to allow wide adoption of this methodology for demanding in vivo and in vitro imaging applications, including cancer research, autoimmune diseases, and tissue engineering., Comment: 20 pages, 5 figures, under review in Science Advances

Based on a mean-field theory of a non-rotating turbulent convection (Phys. Rev. E {\bf 66}, 066305, 2002), we perform mean-field simulations (MFS) of sheared convection which takes into account an effect of modification of the turbulent heat flux by the non-uniform large-scale motions. As the result of this effect, there is an excitation of large-scale convective-shear instability, which causes the formation of large-scale semi-organized structures in the form of rolls. The life-times and spatial scales of these structures are much larger compared to the turbulent scales. By means of MFS performed for stress-free and no-slip vertical boundary conditions, we determine the spatial and temporal characteristics of these structures. Our study demonstrates that the modification of the turbulent heat flux by non-uniform flows leads to a strong reduction of the critical effective Rayleigh number (based on the eddy viscosity and turbulent temperature diffusivity) required for the formation of the large-scale rolls. During the nonlinear stage of the convective-shear instability, there is a transition from the two-layer vertical structure with two roles in the vertical direction before the system reaches steady-state to the one-layer vertical structure with one role after the system reaches steady-state. This effect is observed for all effective Rayleigh numbers. We find that inside the convective rolls, the spatial distribution of the mean potential temperature includes regions with a positive vertical gradient of the potential temperature caused by the mean heat flux of the convective rolls. This study might be useful for understanding of the origin of large-scale rolls observed in atmospheric convective boundary layers as well as in numerical simulations and laboratory experiments., Comment: 13 pages, 17 figures, revtex4-2, revised. arXiv admin note: text overlap with arXiv:2209.07477

For rockfall hazard assessment on areas more than several km2 in size, the quantification of runout probability is usually done empirically. Classical methods use statistical distributions of reach or energy angles derived from rockfall databases. However, other topographic descriptors can be derived from the topographic profiles along the rockfall path. Using a database of more than 4,000 profiles of rockfall paths, we determine which topographic descriptors are most appropriate for reach probability estimation, by comparing their statistical distributions for rockfall stopping points, and for points the rocks overtook. We show that the curvilinear length of the propagation path, and the area under the propagation path can improve propagation estimations, especially when they are computed only along the final portion of the propagation path. This is illustrated by comparing an experimental distribution of rockfall stopping points to the estimated distributions.

We perform in this work an analysis of the background dynamics for $\alpha$-attractor models in the context of loop quantum cosmology. Particular attention is given to the determination of the duration of the inflationary phase that is preceded by the quantum bounce in these models. From an analysis of the general predictions for these models, it is shown that we can be able to put constraints in the parameter $\alpha$ of the potentials and also on the quantum model itself, especially the Barbero-Immirzi parameter. In particular, the constraints on the tensor-to-scalar ratio and spectral tilt of the cosmological perturbations limit the $\alpha$ parameter of the potentials to values such that $\alpha_{n=0} \lesssim 10$, $\alpha_{n=1} \lesssim 17$ and $\alpha_{n=2} \lesssim 67$, for the $\alpha$ attractors T, E, and $n=2$ models, respectively. Using the constraints on the minimal amount of e-folds of expansion from the quantum bounce up to the end of inflation leads to the upper bounds for the Barbero-Immirzi parameter for the $\alpha$-attractor models studied in this work: $\gamma_{n=0} \lesssim 51.2$, $\gamma_{n=1}\lesssim 63.4$ and $\gamma_{n=2} \lesssim 64.2$, which are obtained when fixing the parameter $\alpha$ in the potential at the values saturating the upper bounds given above for each model., Comment: 11 pages, 6 figures. Replaced with the version matching the published one in the Physical Review D

As an organic semiconductor and a prototypical acceptor molecule in organic photovoltaics, C$_{60}$ has broad relevance to the world of organic thin film electronics. Although highly uniform C$_{60}$ thin films are necessary to conduct spectroscopic analysis of the electronic structure of these C$_{60}$-based materials, reported C$_{60}$ films show a relatively low degree of order beyond a monolayer. Here, we develop a generalizable two-stage growth technique that consistently produces single-domain C$_{60}$ films of controllable thicknesses, using Au(111) as an epitaxially well-matched substrate. We characterize the films using low-energy electron diffraction, low-energy electron microscopy, scanning tunneling microscopy, and angle-resolved photoemission spectroscopy (ARPES). We report highly oriented epitaxial film growth of C$_{60}$/Au(111) from 1 monolayer (ML) up to 20 ML films. The high-quality of the C$_{60}$ thin films enables the direct observation of the electronic dispersion of the HOMO and HOMO-1 bands via ARPES without need for small spot sizes. Our results indicate a path for the growth of organic films on metallic substrates with long-range ordering., Comment: 16 pages, 7 figures

The quadratic complexity and weak length extrapolation of Transformers limits their ability to scale to long sequences, and while sub-quadratic solutions like linear attention and state space models exist, they empirically underperform Transformers in pretraining efficiency and downstream task accuracy. We introduce Megalodon, a neural architecture for efficient sequence modeling with unlimited context length. Megalodon inherits the architecture of Mega (exponential moving average with gated attention), and further introduces multiple technical components to improve its capability and stability, including complex exponential moving average (CEMA), timestep normalization layer, normalized attention mechanism and pre-norm with two-hop residual configuration. In a controlled head-to-head comparison with Llama2, Megalodon achieves better efficiency than Transformer in the scale of 7 billion parameters and 2 trillion training tokens. Megalodon reaches a training loss of 1.70, landing mid-way between Llama2-7B (1.75) and 13B (1.67). Code: https://github.com/XuezheMax/megalodon, Comment: 9 pages, 6 figures and 8 tables

We study Monge-Amp\`ere gravity (MAG) as an effective theory of cosmological structure formation through optimal transport theory. MAG is based on the Monge-Amp\`ere equation, a nonlinear version of the Poisson equation, that relates the Hessian determinant of the potential to the density field. We explain how MAG emerges from a conditioned system of independent and indistinguishable Brownian particles, through the large deviation principle, in the continuum limit. To numerically explore this highly non-linear theory, we develop a novel N-body simulation method based on semi-discrete optimal transport. Our results obtained from the very first N-body simulation of Monge-Amp\`ere gravity with over 100 millions particles show that on large scales, Monge-Amp\`ere gravity is similar to the Newtonian gravity but favours the formation of anisotropic structures such as filaments. At small scales, MAG has a weaker clustering and is screened in high-density regions. Although here we study the Monge-Amp\`ere gravity as an effective rather than a fundamental theory, our novel highly-performant optimal transport algorithm can be used to run high-resolution simulations of a large class of modified theories of gravity, such as Galileons, in which the equations of motion are second-order and of Monge-Amp\`ere type., Comment: 16 pages, 10 figures

We measure the stacked lensing signal in the direction of galaxy clusters in the Dark Energy Survey Year 3 (DES Y3) redMaPPer sample, using cosmic microwave background (CMB) temperature data from SPT-3G, the third-generation CMB camera on the South Pole Telescope (SPT). We estimate the lensing signal using temperature maps constructed from the initial 2 years of data from the SPT-3G 'Main' survey, covering 1500 deg$^2$ of the Southern sky. We then use this signal as a proxy for the mean cluster mass of the DES sample. In this work, we employ three versions of the redMaPPer catalogue: a Flux-Limited sample containing 8865 clusters, a Volume-Limited sample with 5391 clusters, and a Volume&Redshift-Limited sample with 4450 clusters. For the three samples, we find the mean cluster masses to be ${M}_{200{\rm{m}}}=1.66\pm0.13$ [stat.]$\pm0.03$ [sys.], $1.97\pm0.18$ [stat.]$\pm0.05$ [sys.], and $2.11\pm0.20$ [stat.]$\pm0.05$ [sys.]$\times{10}^{14}\ {\rm{M}}_{\odot }$, respectively. This is a factor of $\sim2$ improvement relative to the precision of measurements with previous generations of SPT surveys and the most constraining cluster mass measurements using CMB cluster lensing to date. Overall, we find no significant tensions between our results and masses given by redMaPPer mass-richness scaling relations of previous works, which were calibrated using CMB cluster lensing, optical weak lensing, and velocity dispersion measurements from various combinations of DES, SDSS and Planck data. We then divide our sample into 3 redshift and 3 richness bins, finding no significant tensions with optical weak-lensing calibrated masses in these bins. We forecast a $5.7\%$ constraint on the mean cluster mass of the DES Y3 sample with the complete SPT-3G surveys when using both temperature and polarization data and including an additional $\sim1400$ deg$^2$ of observations from the 'Extended' SPT-3G survey., Comment: 23 pages, 9 figures, accepted for publication in JCAP. Minor changes and corrections have been made relative to v1

The LaAlO$_3$/SrTiO$_3$ interface hosts a plethora of gate-tunable electronic phases. Gating of LaAlO$_3$/SrTiO$_3$ interfaces are usually assumed to occur electrostatically. However, increasing evidence suggests that non-local interactions can influence and, in some cases, dominate the coupling between applied gate voltages and electronic properties. Here, we sketch quasi-1D ballistic electron waveguides at the LaAlO$_3$/SrTiO$_3$ interface as a probe to understand how gate tunability varies as a function of spatial separation. Gate tunability measurements reveal the scaling law to be at odds with the pure electrostatic coupling observed in traditional semiconductor systems. The non-Coulombic gating at the interface is attributed to the existence of a long-range nanoelectromechanical coupling between the gate and electron waveguide, mediated by the ferroelastic domains in SrTiO$_3$. The long-range interactions at the LaAlO$_3$/SrTiO$_3$ interface add unexpected richness and complexity to this correlated electron system.

MiMiC is a framework for performing multiscale simulations in which loosely coupled external programs describe individual subsystems at different resolutions and levels of theory. To make it highly efficient and flexible, we adopt an interoperable approach based on a multiple-program multiple-data (MPMD) paradigm, serving as an intermediary responsible for fast data exchange and interactions between the subsystems. The main goal of MiMiC is to avoid interfering with the underlying parallelization of the external programs, including the operability on hybrid architectures (e.g., CPU/GPU), and keep their setup and execution as close as possible to the original. At the moment, MiMiC offers an efficient implementation of electrostatic embedding QM/MM that has demonstrated unprecedented parallel scaling in simulations of large biomolecules using CPMD and GROMACS as QM and MM engines, respectively. However, as it is designed for high flexibility with general multiscale models in mind, it can be straightforwardly extended beyond QM/MM. In this article, we illustrate the software design and the features of the framework, which make it a compelling choice for multiscale simulations in the upcoming era of exascale high-performance computing.

We forecast constraints on cosmological parameters enabled by three surveys conducted with SPT-3G, the third-generation camera on the South Pole Telescope. The surveys cover separate regions of 1500, 2650, and 6000 ${\rm deg}^{2}$ to different depths, in total observing 25% of the sky. These regions will be measured to white noise levels of roughly 2.5, 9, and 12 $\mu{\rm K-arcmin}$, respectively, in CMB temperature units at 150 GHz by the end of 2024. The survey also includes measurements at 95 and 220 GHz, which have noise levels a factor of ~1.2 and 3.5 times higher than 150 GHz, respectively, with each band having a polarization noise level ~$\sqrt{\text{2}}$ times higher than the temperature noise. We use a novel approach to obtain the covariance matrices for jointly and optimally estimated gravitational lensing potential bandpowers and unlensed CMB temperature and polarization bandpowers. We demonstrate the ability to test the $\Lambda{\rm CDM}$ model via the consistency of cosmological parameters constrained independently from SPT-3G and Planck data, and consider the improvement in constraints on $\Lambda{\rm CDM}$ extension parameters from a joint analysis of SPT-3G and Planck data. The $\Lambda{\rm CDM}$ cosmological parameters are typically constrained with uncertainties up to ~2 times smaller with SPT-3G data, compared to Planck, with the two data sets measuring significantly different angular scales and polarization levels, providing additional tests of the standard cosmological model., Comment: 26 pages; 13 figures; Accepted for publication in ApJ; Minor edits have been made

Grounded text generation, encompassing tasks such as long-form question-answering and summarization, necessitates both content selection and content consolidation. Current end-to-end methods are difficult to control and interpret due to their opaqueness. Accordingly, recent works have proposed a modular approach, with separate components for each step. Specifically, we focus on the second subtask, of generating coherent text given pre-selected content in a multi-document setting. Concretely, we formalize Fusion-in-Context (FiC) as a standalone task, whose input consists of source texts with highlighted spans of targeted content. A model then needs to generate a coherent passage that includes all and only the target information. Our work includes the development of a curated dataset of 1000 instances in the reviews domain, alongside a novel evaluation framework for assessing the faithfulness and coverage of highlights, which strongly correlate to human judgment. Several baseline models exhibit promising outcomes and provide insightful analyses. This study lays the groundwork for further exploration of modular text generation in the multi-document setting, offering potential improvements in the quality and reliability of generated content. Our benchmark, FuseReviews, including the dataset, evaluation framework, and designated leaderboard, can be found at https://fusereviews.github.io/., Comment: NAACL 2024, findings

Underwater image restoration is a challenging task because of water effects that increase dramatically with distance. This is worsened by lack of ground truth data of clean scenes without water. Diffusion priors have emerged as strong image restoration priors. However, they are often trained with a dataset of the desired restored output, which is not available in our case. We also observe that using only color data is insufficient, and therefore augment the prior with a depth channel. We train an unconditional diffusion model prior on the joint space of color and depth, using standard RGBD datasets of natural outdoor scenes in air. Using this prior together with a novel guidance method based on the underwater image formation model, we generate posterior samples of clean images, removing the water effects. Even though our prior did not see any underwater images during training, our method outperforms state-of-the-art baselines for image restoration on very challenging scenes. Our code, models and data are available on the project website., Comment: ECCV 2024. Project page with results and code: https://osmosis-diffusion.github.io/

Motivated by recent experiments (Subero et. al. Nature Comm. $\bf{14}$, 7924 (2023)) we analyze photonic heat transport through a Josephson junction in a dissipative environment. For this purpose we derive the general expressions for the heat current in terms of non-equilibrium Green functions for the junction coupled in series or in parallel with two environmental impedances at different temperatures. We show that even on the insulating side of the Schmid transition the heat current is sensitive to the Josephson coupling exhibiting an opposite behavior for the series and parallel connection and in qualitative agreement with experiments. We also predict that this device should exhibit heat rectification properties and provide simple expressions to account for them in terms of the system parameters.

Neural Radiance Fields (NeRFs) are trained to minimize the rendering loss of predicted viewpoints. However, the photometric loss often does not provide enough information to disambiguate between different possible geometries yielding the same image. Previous work has thus incorporated depth supervision during NeRF training, leveraging dense predictions from pre-trained depth networks as pseudo-ground truth. While these depth priors are assumed to be perfect once filtered for noise, in practice, their accuracy is more challenging to capture. This work proposes a novel approach to uncertainty in depth priors for NeRF supervision. Instead of using custom-trained depth or uncertainty priors, we use off-the-shelf pretrained diffusion models to predict depth and capture uncertainty during the denoising process. Because we know that depth priors are prone to errors, we propose to supervise the ray termination distance distribution with Earth Mover's Distance instead of enforcing the rendered depth to replicate the depth prior exactly through L2-loss. Our depth-guided NeRF outperforms all baselines on standard depth metrics by a large margin while maintaining performance on photometric measures., Comment: Accepted to ECCV 2024

Advancements in vision-language models (VLMs) have propelled the field of computer vision, particularly in the zero-shot learning setting. Despite their promise, the effectiveness of these models often diminishes due to domain shifts in test environments. To address this, we introduce the Test-Time Prototype Shifting (TPS) framework, a pioneering approach designed to adapt VLMs to test datasets using unlabeled test inputs. Our method is based on the notion of modulating per-class prototypes in the shared embedding space. By pre-computing and caching prototypes generated with the pre-trained text encoder, TPS not only facilitates optimization-free prototype reuse for subsequent predictions but also enables seamless integration with current advancements in prompt engineering. At test-time, TPS dynamically learns shift vectors for each prototype based solely on the given test sample, effectively bridging the domain gap and enhancing classification accuracy. A notable aspect of our framework is its significantly reduced memory and computational demands when compared to conventional text-prompt tuning methods. Extensive evaluations across 15 datasets involving natural distribution shifts and cross-dataset generalization demonstrate TPS's superior performance, achieving state-of-the-art results while reducing resource requirements.

We demonstrate the effectiveness of one of the many multi-tracer analyses enabled by Optimal Transport (OT) reconstruction. Leveraging a semi-discrete OT algorithm, we determine the displacements between initial and observed positions of biased tracers and the remaining matter field. With only redshift-space distorted final positions of biased tracers and a simple premise for the remaining mass distribution as input, OT solves the displacement field. This extracted field, assuming asymptotically uniform density and a gradient flow displacement, enables reconstruction of the initial overdensity fluctuation field. We show that the divergence of the OT displacement field is a good proxy of the linear density field, even though the method never assumes the linear theory growth. Additionally, this divergence field can be combined with the reconstructed protohalos to provide a higher signal-to-noise measurement of the BAO standard ruler than was possible with either measurement individually., Comment: 12 pages, 10 figures

Benchmarks of the multilingual capabilities of text-to-image (T2I) models compare generated images prompted in a test language to an expected image distribution over a concept set. One such benchmark, "Conceptual Coverage Across Languages" (CoCo-CroLa), assesses the tangible noun inventory of T2I models by prompting them to generate pictures from a concept list translated to seven languages and comparing the output image populations. Unfortunately, we find that this benchmark contains translation errors of varying severity in Spanish, Japanese, and Chinese. We provide corrections for these errors and analyze how impactful they are on the utility and validity of CoCo-CroLa as a benchmark. We reassess multiple baseline T2I models with the revisions, compare the outputs elicited under the new translations to those conditioned on the old, and show that a correction's impactfulness on the image-domain benchmark results can be predicted in the text domain with similarity scores. Our findings will guide the future development of T2I multilinguality metrics by providing analytical tools for practical translation decisions., Comment: NAACL 2024 Main Conference

Long-form video understanding represents a significant challenge within computer vision, demanding a model capable of reasoning over long multi-modal sequences. Motivated by the human cognitive process for long-form video understanding, we emphasize interactive reasoning and planning over the ability to process lengthy visual inputs. We introduce a novel agent-based system, VideoAgent, that employs a large language model as a central agent to iteratively identify and compile crucial information to answer a question, with vision-language foundation models serving as tools to translate and retrieve visual information. Evaluated on the challenging EgoSchema and NExT-QA benchmarks, VideoAgent achieves 54.1% and 71.3% zero-shot accuracy with only 8.4 and 8.2 frames used on average. These results demonstrate superior effectiveness and efficiency of our method over the current state-of-the-art methods, highlighting the potential of agent-based approaches in advancing long-form video understanding.

The ability to accurately infer cardiac electrophysiological (EP) properties is key to improving arrhythmia diagnosis and treatment. In this work, we developed a physics-informed neural networks (PINNs) framework to predict how different myocardial EP parameters are modulated by anti-arrhythmic drugs. Using $\textit{in vitro}$ optical mapping images and the 3-channel Fenton-Karma model, we estimated the changes in ionic channel conductance caused by these drugs. Our framework successfully characterised the action of drugs HMR1556, nifedipine and lidocaine - respectively, blockade of $I_{K}$, $I_{Ca}$, and $I_{Na}$ currents - by estimating that they decreased the respective channel conductance by $31.8\pm2.7\%$ $(p=8.2 \times 10^{-5})$, $80.9\pm21.6\%$ $(p=0.02)$, and $8.6\pm0.5\%$ $ (p=0.03)$, leaving the conductance of other channels unchanged. For carbenoxolone, whose main action is the blockade of intercellular gap junctions, PINNs also successfully predicted no significant changes $(p>0.09)$ in all ionic conductances. Our results are an important step towards the deployment of PINNs for model parameter estimation from experimental data, bringing this framework closer to clinical or laboratory images analysis and for the personalisation of mathematical models., Comment: Accepted for publication in the 21st IEEE International Symposium on Biomedical Imaging 2024

The scaling laws and extraordinary performance of large foundation models motivate the development and utilization of such models in biomedicine. However, despite early promising results on some biomedical benchmarks, there are still major challenges that need to be addressed before these models can be used in real-world clinics. Frontier general-domain models such as GPT-4V still have significant performance gaps in multimodal biomedical applications. More importantly, less-acknowledged pragmatic issues, including accessibility, model cost, and tedious manual evaluation make it hard for clinicians to use state-of-the-art large models directly on private patient data. Here, we explore training open-source small multimodal models (SMMs) to bridge competency gaps for unmet clinical needs in radiology. To maximize data efficiency, we adopt a modular approach by incorporating state-of-the-art pre-trained models for image and text modalities, and focusing on training a lightweight adapter to ground each modality to the text embedding space, as exemplified by LLaVA-Med. For training, we assemble a large dataset of over 697 thousand radiology image-text pairs. For evaluation, we propose CheXprompt, a GPT-4-based metric for factuality evaluation, and demonstrate its parity with expert evaluation. For best practice, we conduct a systematic ablation study on various choices in data engineering and multimodal training. The resulting LlaVA-Rad (7B) model attains state-of-the-art results on standard radiology tasks such as report generation and cross-modal retrieval, even outperforming much larger models such as GPT-4V and Med-PaLM M (84B). The inference of LlaVA-Rad is fast and can be performed on a single V100 GPU in private settings, offering a promising state-of-the-art tool for real-world clinical applications.

The finding of small price changes in many retail price datasets is often viewed as a puzzle. We show that a possible explanation for the presence of small price changes is related to sales volume, an observation that has been overlooked in the existing literature. Analyzing a large retail scanner price dataset that contains information on both prices and sales volume, we find that small price changes are more frequent when products sales volume is high. This finding holds across product categories, within product categories, and for individual products. It is also robust to various sensitivity analyses such as measurement errors, the definition of small price changes, the inclusion of measures of price synchronization, the size of producers, the time horizon used to compute the average sales volume, the revenues, the competition, shoppers characteristics, etc.

We present the first finite time global convergence analysis of policy gradient in the context of infinite horizon average reward Markov decision processes (MDPs). Specifically, we focus on ergodic tabular MDPs with finite state and action spaces. Our analysis shows that the policy gradient iterates converge to the optimal policy at a sublinear rate of $O\left({\frac{1}{T}}\right),$ which translates to $O\left({\log(T)}\right)$ regret, where $T$ represents the number of iterations. Prior work on performance bounds for discounted reward MDPs cannot be extended to average reward MDPs because the bounds grow proportional to the fifth power of the effective horizon. Thus, our primary contribution is in proving that the policy gradient algorithm converges for average-reward MDPs and in obtaining finite-time performance guarantees. In contrast to the existing discounted reward performance bounds, our performance bounds have an explicit dependence on constants that capture the complexity of the underlying MDP. Motivated by this observation, we reexamine and improve the existing performance bounds for discounted reward MDPs. We also present simulations to empirically evaluate the performance of average reward policy gradient algorithm., Comment: 29 pages, 5 figures