Search

The proliferation of AI-generated images has intensified the need for robust content authentication methods. We present InvisMark, a novel watermarking technique designed for high-resolution AI-generated images. Our approach leverages advanced neural network architectures and training strategies to embed imperceptible yet highly robust watermarks. InvisMark achieves state-of-the-art performance in imperceptibility (PSNR$\sim$51, SSIM $\sim$ 0.998) while maintaining over 97\% bit accuracy across various image manipulations. Notably, we demonstrate the successful encoding of 256-bit watermarks, significantly expanding payload capacity while preserving image quality. This enables the embedding of UUIDs with error correction codes, achieving near-perfect decoding success rates even under challenging image distortions. We also address potential vulnerabilities against advanced attacks and propose mitigation strategies. By combining high imperceptibility, extended payload capacity, and resilience to manipulations, InvisMark provides a robust foundation for ensuring media provenance in an era of increasingly sophisticated AI-generated content. Source code of this paper is available at: https://github.com/microsoft/InvisMark.

We begin by reexamining the holographic reconstruction of scalar fields in four-dimensional anti-de Sitter spacetime, adopting a purely Lorentzian signature derivation, reproducing earlier results of HKLL and generalizing to arbitrary boundary metrics. The approach is extended to gravitational perturbations, focussing on perturbations around $AdS_{4}$ and show that the mapping can be formulated as a purely light-like integral of the conformal field theory stress energy tensor. An example is considered of relevance to the flat spacetime limit with nontrivial BMS charges turned on, potentially providing a quantum field theory definition of celestial CFT as a large central charge limit of a 3d CFT., Comment: 12 pages, 1 figure

Tree tensor network states (TTNS) generalize the notion of having low Schmidt-rank to multipartite quantum states, through a parameter known as the bond dimension. This leads to succinct representations of quantum many-body systems with a tree-like entanglement structure. In this work, we study the task of testing whether an unknown pure state is a TTNS on $n$ qudits with bond dimension at most $r$, or is far in trace distance from any such state. We first establish that, independent of the physical dimensions, $O(nr^2)$ copies of the state suffice to acccomplish this task with one-sided error, as in the matrix product state (MPS) test of Soleimanifar and Wright. We then prove that $\Omega(n r^2/\log n)$ copies are necessary for any test with one-sided error whenever $r\geq 2 + \log n$. In particular, this closes a quadratic gap in the previous bounds for MPS testing in this setting, up to log factors. On the other hand, when $r=2$ we show that $\Theta(\sqrt{n})$ copies are both necessary and sufficient for the related task of testing whether a state is a product of $n$ bipartite states having Schmidt-rank at most $r$, for some choice of physical dimensions. We also study the performance of tests using measurements performed on a small number of copies at a time. In the setting of one-sided error, we prove that adaptive measurements offer no improvement over non-adaptive measurements for many properties, including TTNS. We then derive a closed-form solution for the acceptance probability of an $(r+1)$-copy rank test with rank parameter $r$. This leads to nearly tight bounds for testing rank, Schmidt-rank, and TTNS when the tester is restricted to making measurements on $r+1$ copies at a time. For example, when $r=2$ there is an MPS test with one-sided error which uses $O(n^2)$ copies and measurements performed on just three copies at a time., Comment: 47 pages, 3 figures

Quantum game theory has emerged as a promising candidate to further the understanding of quantum correlations. Motivated by this, it is demonstrated that pure strategy Nash equilibria can be utilised as a mechanism to witness and determine quantum correlation. By combining quantum theory with Bayesian game theory, a constant-sum game is designed in which the players are competing against each other, and crucially gain at the other player's expense. Subsequently, it is found that mixed strategy Nash equilibria are only necessary when considering quantum correlation for the designed game. This reveals that a Bayesian game-theoretic framework yields a sufficient condition in which to detect quantum effects.

M dwarf systems offer a unique opportunity to study terrestrial exoplanetary atmospheres due to their smaller size and cooler temperatures. However, due to the extreme conditions these host stars impose, it is unclear whether their small, close-in rocky planets are able to retain any atmosphere at all. The Hot Rocks Survey aims to answer this question by targeting nine different M dwarf rocky planets spanning a range of planetary and stellar properties. LHS 1478 b orbits an M3-type star, has an equilibrium temperature of Teq = 585 K and experiences an instellation 21 times greater than that of Earth. We observe two secondary eclipses using photometric imaging at 15 um using the Mid-Infrared Instrument on the James Webb Space Telescope (JWST MIRI) to measure thermal emission from the dayside of the planet. We then compare these values to different atmospheric scenarios to evaluate potential heat transport and CO2 absorption signatures. We find a secondary eclipse depth of 146 +/- 56 ppm based on the first observation, while the second observation results in a non-detection due to significantly larger unexplained systematics. Based on the first observation alone, we can reject the null hypothesis of the dark (zero Bond albedo) no atmosphere bare rock model with a confidence level of 3.4 sigma. For an airless body with a Bond albedo of A=0.2, the significance decreases to 2.9 sigma. The secondary eclipse depth is consistent with the majority of atmospheric scenarios we considered, which all involve atmospheres which include different concentrations of CO2, and surface pressures from 0.1 to 10 bar. However, we stress that the two observations from our program do not yield consistent results, and more observations are needed to verify our findings. The Hot Rocks Survey serves as a relevant primer for future endeavors such as the Director's Discretionary Time (DDT) Rocky Worlds program., Comment: Submitted to A&A

The nearby transiting rocky exoplanet LTT 1445A b presents an ideal target for studying atmospheric retention in terrestrial planets orbiting M dwarfs. It is cooler than many rocky exoplanets yet tested for atmospheres, receiving a bolometric instellation similar to Mercury's. Previous transmission spectroscopy ruled out a light H/He-dominated atmosphere but could not distinguish between a bare-rock, a high-MMW, or a cloudy atmosphere. We present new secondary eclipse observations using JWST's MIRI/LRS, covering the 5-12 $\mu$m range. From these observations, we detect a broadband secondary eclipse depth of 41 $\pm$ 9 ppm and measure a mid-eclipse timing consistent with a circular orbit (at 1.7$\sigma$). From its emission spectrum, the planet's dayside brightness temperature is constrained to 525 $\pm$ 15 K, yielding a temperature ratio relative to the maximum average dayside temperature from instant thermal reradiation by a rocky surface $R$ = $T_{\rm day,obs}/T_{\rm max}$ = 0.952 $\pm$ 0.057, consistent with emission from a dark rocky surface. From an energy balance perspective, such a warm dayside temperature disfavors thick atmospheres, excluding $\sim$100 bar atmospheres with Bond albedo $>$ 0.08 at the 3$\sigma$ level. Furthermore, forward modeling of atmospheric emission spectra disfavor simple 100\% CO$_2$ atmospheres with surface pressures of 1, 10, and 100 bar at 4.2$\sigma$, 6.6$\sigma$, and 6.8$\sigma$ confidence, respectively. These results suggest that LTT 1445A b lacks a very thick CO$_2$ atmosphere, possibly due to atmospheric erosion driven by stellar activity. However, the presence of a moderately thin atmosphere (similar to those on Mars, Titan, or Earth) remains uncertain., Comment: 31 pages, 14 figures, submitted to AJ

In recent years, machine learning models have made strides towards human-like reasoning capabilities from several directions. In this work, we review the current state of the literature and describe the remaining steps to achieve a neural model which can perform System~2 reasoning analogous to a human. We argue that if current models are insufficient to be classed as performing reasoning, there remains very little additional progress needed to attain that goal.

Empowering people who are blind or visually impaired (BVI) to enhance their orientation and mobility skills is critical to equalizing their access to social and economic opportunities. To manage this crucial challenge, we employed a novel design process based on a large-scale, longitudinal, community-based structure. Across three annual programs we engaged with the BVI community in online and in-person modes. In total, our team included 67 total BVI participatory design participants online, 11 BVI co-designers in-person, and 4 BVI program coordinators. Through this design process we built a mobile application that enables users to generate, share, and navigate maps of indoor and outdoor environments without the need to instrument each environment with beacons or fiducial markers. We evaluated this app at a healthcare facility, and participants in the evaluation rated the app highly with respect to its design, features, and potential for positive impact on quality of life.

Fast X-ray transients (FXTs) are extragalactic bursts of X-rays first identified in archival X-ray data, and now routinely discovered by the Einstein Probe in real time, which is continuously surveying the night sky in the soft ($0.5 - 4$ keV) X-ray regime. In this Letter, we report the discovery of the second optical counterpart (AT2024gsa) to an FXT (EP240414a). EP240414a is located at a projected radial separation of 27 kpc from its likely host galaxy at $z = 0.4018 \pm 0.0010$. The optical light curve of AT2024gsa displays three distinct components. The initial decay from our first observation is followed by a re-brightening episode, displaying a rapid rise in luminosity to an absolute magnitude of $M_r \sim -21$ after two rest-frame days. While the early optical luminosity and decline rate is similar to luminous fast blue optical transients, the colour temperature of AT2024gsa is distinctly red and we show that the peak flux is inconsistent with a thermal origin. The third component peaks at $M_i \sim -19$ at $\gtrsim 16$ rest-frame days post-FXT, and is compatible with an emerging supernova. We fit the $riz$-band data with a series of power laws and find that the decaying components are in agreement with gamma-ray burst afterglow models, and that the re-brightening may originate from refreshed shocks. By considering EP240414a in context with all previously reported known-redshift FXT events, we propose that Einstein Probe FXT discoveries may all result from high-redshift gamma-ray bursts, and thus are distinct from the previously discovered lower redshift, lower luminosity population of FXTs., Comment: To be submitted to ApJL

The strong electron-electron interaction in transition metal dichalcogenides (TMDs) gives rise to phenomena such as strong exciton and trion binding and excitonic condensation, as well as large negative exchange and correlation contributions to the electron energies, resulting in negative electronic compressibility. Here we use angle-resolved photoemission spectroscopy to demonstrate a striking effect of negative electronic compressibility in semiconducting TMD MoS2 on the charge transfer to and from a partial overlayer of monolayer semimetallic WTe2. We track the changes in binding energy of the valence bands of both WTe2 and MoS2 as a function of surface transfer doping with donor (K) and acceptor (F4-TCNQ) species. Donor doping increases the binding energy of the MoS2 valence band, as expected, while counterintuitively reducing the binding energy of the WTe2 valence bands and core levels. The inverse effect is observed for acceptor doping, where a typical reduction in the MoS2 binding energies is accopanied by an unexpected increase in those of WTe2. The observations imply a reversal of the expected charge transfer; donor (acceptor) deposition decreases (increases) the carrier density in the WTe2 adlayer. The charge transfer reversal is a direct consequence of the negative electronic compressibility of the MoS2 surface layer, for which addition (subtraction) of charge leads to attraction (repulsion) of further charge from neighbouring layers. These findings highlight the importance of many-body interactions for the electrons in transition metal dichalcogenides and underscore the potential for exploring strongly correlated quantum states in two-dimensional semiconductors., Comment: 4 pages, 5 figures + supplementary information

In this work, we apply state-of-the-art self-supervised learning techniques on a large dataset of seafloor imagery, \textit{BenthicNet}, and study their performance for a complex hierarchical multi-label (HML) classification downstream task. In particular, we demonstrate the capacity to conduct HML training in scenarios where there exist multiple levels of missing annotation information, an important scenario for handling heterogeneous real-world data collected by multiple research groups with differing data collection protocols. We find that, when using smaller one-hot image label datasets typical of local or regional scale benthic science projects, models pre-trained with self-supervision on a larger collection of in-domain benthic data outperform models pre-trained on ImageNet. In the HML setting, we find the model can attain a deeper and more precise classification if it is pre-trained with self-supervision on in-domain data. We hope this work can establish a benchmark for future models in the field of automated underwater image annotation tasks and can guide work in other domains with hierarchical annotations of mixed resolution.

Self-supervised learning (SSL) is an effective method for exploiting unlabelled data to learn a high-level embedding space that can be used for various downstream tasks. However, existing methods to monitor the quality of the encoder -- either during training for one model or to compare several trained models -- still rely on access to annotated data. When SSL methodologies are applied to new data domains, a sufficiently large labelled dataset may not always be available. In this study, we propose several evaluation metrics which can be applied on the embeddings of unlabelled data and investigate their viability by comparing them to linear probe accuracy (a common metric which utilizes an annotated dataset). In particular, we apply $k$-means clustering and measure the clustering quality with the silhouette score and clustering agreement. We also measure the entropy of the embedding distribution. We find that while the clusters did correspond better to the ground truth annotations as training of the network progressed, label-free clustering metrics correlated with the linear probe accuracy only when training with SSL methods SimCLR and MoCo-v2, but not with SimSiam. Additionally, although entropy did not always have strong correlations with LP accuracy, this appears to be due to instability arising from early training, with the metric stabilizing and becoming more reliable at later stages of learning. Furthermore, while entropy generally decreases as learning progresses, this trend reverses for SimSiam. More research is required to establish the cause for this unexpected behaviour. Lastly, we find that while clustering based approaches are likely only viable for same-architecture comparisons, entropy may be architecture-independent.

The EXperiment for Cryogenic Large-Aperture Intensity Mapping (EXCLAIM) is a balloon-borne telescope designed to survey star formation over cosmological time scales using intensity mapping in the 420 - 540 GHz frequency range. EXCLAIM uses a fully cryogenic telescope coupled to six on-chip spectrometers featuring kinetic inductance detectors (KIDs) to achieve high sensitivity, allowing for fast integration in dark atmospheric windows. The telescope receiver is cooled to $\approx$ 1.7 K by immersion in a superfluid helium bath and enclosed in a superfluid-tight shell with a meta-material anti-reflection coated silicon window. In addition to the optics and the spectrometer package, the receiver contains the magnetic shielding, the cryogenic segment of the spectrometer readout, and the sub-Kelvin cooling system. A three-stage continuous adiabatic demagnetization refrigerator (CADR) keeps the detectors at 100 mK while a $^4$He sorption cooler provides a 900 mK thermal intercept for mechanical suspensions and coaxial cables. We present the design of the EXCLAIM receiver and report on the flight-like testing of major receiver components, including the superfluid-tight receiver window and the sub-Kelvin coolers., Comment: 13 pages, 7 figures, 1 table

Quasi-periodic Eruptions (QPEs) are luminous bursts of soft X-rays from the nuclei of galaxies, repeating on timescales of hours to weeks. The mechanism behind these rare systems is uncertain, but most theories involve accretion disks around supermassive black holes (SMBHs), undergoing instabilities or interacting with a stellar object in a close orbit. It has been suggested that this disk could be created when the SMBH disrupts a passing star, implying that many QPEs should be preceded by observable tidal disruption events (TDEs). Two known QPE sources show long-term decays in quiescent luminosity consistent with TDEs, and two observed TDEs have exhibited X-ray flares consistent with individual eruptions. TDEs and QPEs also occur preferentially in similar galaxies. However, no confirmed repeating QPEs have been associated with a spectroscopically confirmed TDE or an optical TDE observed at peak brightness. Here we report the detection of nine X-ray QPEs with a mean recurrence time of approximately 48 hours from AT2019qiz, a nearby and extensively studied optically-selected TDE. We detect and model the X-ray, ultraviolet and optical emission from the accretion disk, and show that an orbiting body colliding with this disk provides a plausible explanation for the QPEs.

Directional wave speeds variations in anisotropic elastic solids enables material characterisation capabilities, such as determination of elastic constants and volumetric measurement of crystallographic texture. However, achieving such measurements is challenging especially on samples with complex geometries. Here we propose the use of Green's Function reconstruction from diffuse ultrasonic wave fields for accurate velocity measurements on components with arbitrary geometries. Strategies for accurate reconstruction, including averaging over an increased number of different source locations, using longer window lengths of diffuse fields, and accurately deconvolving a source-dependent factor, were implemented to achieve satisfactory convergence towards Green's Function. Additionally, low signal intensity challenges from laser interferometers were overcome to enable non-contact measurement of the wave speeds, by making use of simultaneous excitation of sources to increase signal-to-noise ratio and signal normalisation to account for energy dissipation of diffuse fields. With successful demonstration using both phased array and laser receivers, this advancement fundamentally broadens acoustic wave velocity measurement capabilities to a wider range of environments and holds promise for future material characterisation of complex-shaped components.

We prove that a closed negatively curved analytic Riemannian manifold that contains infinitely many totally geodesic hypersurfaces is isometric to an arithmetic hyperbolic manifold. Equivalently, any closed analytic Riemannian manifold with negative sectional curvature has only finitely many totally geodesic hypersurfaces, unless it has constant curvature., Comment: 30 pages, minor revisions

This paper explores the potential of cryogenic semiconductor computing and superconductor electronics as promising alternatives to traditional semiconductor devices. As semiconductor devices face challenges such as increased leakage currents and reduced performance at higher temperatures, these novel technologies offer high performance and low power computation. Conventional semiconductor electronics operating at cryogenic temperatures (below -150{\deg}C or 123.15 K) can benefit from reduced leakage currents and improved electron mobility. On the other hand, superconductor electronics, operating below 10 K, allow electrons to flow without resistance, offering the potential for ultra-low-power, high-speed computation. This study presents a comprehensive performance modeling and analysis of these technologies and provides insights into their potential benefits and limitations. We implement models of in-order and out-of-order cores operating at high clock frequencies associated with superconductor electronics and cryogenic semiconductor computing in gem5. We evaluate the performance of these components using workloads representative of real-world applications like NPB, SPEC CPU2006, and GAPBS. Our results show the potential speedups achievable by these components and the limitations posed by cache bandwidth. This work provides valuable insights into the performance implications and design trade-offs associated with cryogenic and superconductor technologies, laying the foundation for future research in this field using gem5.

In this work, we introduce Gemma 2, a new addition to the Gemma family of lightweight, state-of-the-art open models, ranging in scale from 2 billion to 27 billion parameters. In this new version, we apply several known technical modifications to the Transformer architecture, such as interleaving local-global attentions (Beltagy et al., 2020a) and group-query attention (Ainslie et al., 2023). We also train the 2B and 9B models with knowledge distillation (Hinton et al., 2015) instead of next token prediction. The resulting models deliver the best performance for their size, and even offer competitive alternatives to models that are 2-3 times bigger. We release all our models to the community.

Witnessing quantum correlation in real, practical experimental set-ups is a key focus for the development of near-future quantum technologies. Whilst there are several experimental protocols for witnessing entanglement, detecting quantum discord remains far more elusive. A recently proposed quantum discord witness offers an experimentally accessible set-up which allows the quantum discord to be witnessed through a non-linear combination of correlation functions. Interestingly, the experimental set-up can be mapped to Bayesian game theory allowing for an extended generalisation of the proposed witness. Subsequently, it is shown that there is a direct link between the expected payoff in Bayesian game theory and the previously proposed quantum discord witness by uniting these two concepts through the established CHSH game.

Machine learning models are exceptionally effective in capturing complex non-linear relationships of high-dimensional datasets and making accurate predictions. However, their intrinsic ``black-box'' nature makes it difficult to interpret them or guarantee ``safe behavior'' when deployed in high-risk applications such as feedback control, healthcare and finance. This drawback acts as a significant barrier to their wider application across many scientific and industrial domains where the interpretability of the model predictions is as important as accuracy. Leveraging the latest developments in interpretable machine learning, we develop a method to parameterise ``black-box'' models, effectively transforming them into ``grey-box'' models. We apply this approach to plasma diagnostics by creating a parameterised synthetic Soft X-Ray imaging $-$ Thomson Scattering diagnostic, which predicts high temporal resolution electron temperature and density profiles from the measured soft X-ray emission. The ``grey-box'' model predictions are benchmarked against the trained ``black-box'' models as well as a diverse range of plasma conditions. Our model-agnostic approach can be applied to various machine learning architectures, enabling direct comparisons of model interpretations., Comment: 32 pages, 15 figures

We report on the effects of cosmic ray interactions with the Kinetic Inductance Detector (KID) based focal plane array for the Terahertz Intensity Mapper (TIM). TIM is a NASA-funded balloon-borne experiment designed to probe the peak of the star formation in the Universe. It employs two spectroscopic bands, each equipped with a focal plane of four $\sim\,$900-pixel, KID-based array chips. Measurements of an 864-pixel TIM array shows 791 resonators in a 0.5$\,$GHz bandwidth. We discuss challenges with resonator calibration caused by this high multiplexing density. We robustly identify the physical positions of 788 (99.6$\,$%) detectors using a custom LED-based identification scheme. Using this information we show that cosmic ray events occur at a rate of 2.1$\,\mathrm{events/min/cm^2}$ in our array. 66$\,$% of the events affect a single pixel, and another 33$\,$% affect $<\,$5 KIDs per event spread over a 0.66$\,\mathrm{cm^2}$ region (2 pixel pitches in radius). We observe a total cosmic ray dead fraction of 0.0011$\,$%, and predict that the maximum possible in-flight dead fraction is $\sim\,$0.165$\,$%, which demonstrates our design will be robust against these high-energy events., Comment: 14 pages, 5 figures. Accepted for the publication in Journal of Low Temperature Physics (2024)

We present a statistical modelling framework for implementing Distributed Lag Models (DLMs), encompassing several extensions of the approach to capture the temporally distributed effect from covariates via regression. We place DLMs in the context of penalised Generalized Additive Models (GAMs) and illustrate that implementation via the R package \texttt{mgcv}, which allows for flexible and interpretable inference in addition to thorough model assessment. We show how the interpretation of penalised splines as random quantities enables approximate Bayesian inference and hierarchical structures in the same practical setting. We focus on epidemiological studies and demonstrate the approach with application to mortality data from Cyprus and Greece. For the Cyprus case study, we investigate for the first time, the joint lagged effects from both temperature and humidity on mortality risk with the unexpected result that humidity severely increases risk during cold rather than hot conditions. Another novel application is the use of the proposed framework for hierarchical pooling, to estimate district-specific covariate-lag risk on morality and the use of posterior simulation to compare risk across districts., Comment: Submitted for review in "Statistics in Medicine" on 4/January/ 2024

The Hamiltonian governing the gravitational interaction of N relativistic particles in a four-dimensional anti-de Sitter background is derived to leading order in Newton's constant. The resulting pairwise interactions, combined with the confining nature of motion in anti-de Sitter spacetime, are expected to lead to classical chaos. In the context of the AdS/CFT correspondence, the emergence of a chaotic classical limit on the gravity side has important implications for the dual three-dimensional conformal field theory, including that the spectrum of conformal primary operators at strong coupling should exhibit level repulsion in line with the Wigner surmise., Comment: 18 pages

We propose to build a Flexible Stellarator Physics Facility to explore promising regions of the vast parameter space of disruption-free stellarator solutions for Fusion Pilot Plants (FPPs)., Comment: White paper submitted to FESAC subcommittee on Facilities, 8 pages

Owing to the recent success of Large Language Models, Modern A.I has been much focused on linguistic interactions with humans but less focused on non-linguistic forms of communication between man and machine. In the present paper, we test how affective-linguistic communication, in combination with differential outcomes training, affects mutual learning in a human-robot context. Taking inspiration from child-caregiver dynamics, our human-robot interaction setup consists of a (simulated) robot attempting to learn how best to communicate internal, homeostatically-controlled needs; while a human "caregiver" attempts to learn the correct object to satisfy the robot's present communicated need. We studied the effects of i) human training type, and ii) robot reinforcement learning type, to assess mutual learning terminal accuracy and rate of learning (as measured by the average reward achieved by the robot). Our results find mutual learning between a human and a robot is significantly improved with Differential Outcomes Training (DOT) compared to Non-DOT (control) conditions. We find further improvements when the robot uses an exploration-exploitation policy selection, compared to purely exploitation policy selection. These findings have implications for utilizing socially assistive robots (SAR) in therapeutic contexts, e.g. for cognitive interventions, and educational applications., Comment: 14 pages, with references; 1 figure, 3 tables

The high-energy radiative output, from the X-ray to the ultraviolet, of exoplanet host stars drives photochemical reactions and mass loss in the upper regions of planetary atmospheres. In order to place constraints on the atmospheric properties of the three closest terrestrial exoplanets transiting M dwarfs, we observe the high-energy spectra of the host stars LTT1445A and GJ486 in the X-ray with XMM-Newton and Chandra and in the ultraviolet with HST/COS and STIS. We combine these observations with estimates of extreme ultraviolet flux, reconstructions of the Ly-a lines, and stellar models at optical and infrared wavelengths to produce panchromatic spectra from 1A--20um for each star. While LTT1445Ab, LTT1445Ac, and GJ486b do not possess primordial hydrogen-dominated atmospheres, we calculate that they are able to retain pure CO2 atmospheres if starting with 10, 15, and 50% of Earth's total CO2 budget, respectively, in the presence of their host stars' stellar wind. We use age-activity relationships to place lower limits of 2.2 and 6.6 Gyr on the ages of the host stars LTT1445A and GJ486. Despite both LTT1445A and GJ486 appearing inactive at optical wavelengths, we detect flares at ultraviolet and X-ray wavelengths for both stars. In particular, GJ486 exhibits two flares with absolute energies of 10^29.5 and 10^30.1 erg (equivalent durations of 4357+/-96 and 19724+/-169 s) occurring three hours apart, captured with HST/COS G130M. Based on the timing of the observations, we suggest that these high-energy flares are related and indicative of heightened flaring activity that lasts for a period of days, but our interpretations are limited by sparse time-sampling. Consistent high-energy monitoring is needed to determine the duration and extent of high-energy activity on individual M dwarfs, as well as the population as a whole., Comment: 21 pages, published in A&A

As part of an ongoing worldwide effort to comprehend and monitor insect biodiversity, this paper presents the BIOSCAN-5M Insect dataset to the machine learning community and establish several benchmark tasks. BIOSCAN-5M is a comprehensive dataset containing multi-modal information for over 5 million insect specimens, and it significantly expands existing image-based biological datasets by including taxonomic labels, raw nucleotide barcode sequences, assigned barcode index numbers, geographical, and size information. We propose three benchmark experiments to demonstrate the impact of the multi-modal data types on the classification and clustering accuracy. First, we pretrain a masked language model on the DNA barcode sequences of the BIOSCAN-5M dataset, and demonstrate the impact of using this large reference library on species- and genus-level classification performance. Second, we propose a zero-shot transfer learning task applied to images and DNA barcodes to cluster feature embeddings obtained from self-supervised learning, to investigate whether meaningful clusters can be derived from these representation embeddings. Third, we benchmark multi-modality by performing contrastive learning on DNA barcodes, image data, and taxonomic information. This yields a general shared embedding space enabling taxonomic classification using multiple types of information and modalities. The code repository of the BIOSCAN-5M Insect dataset is available at https://github.com/bioscan-ml/BIOSCAN-5M.

Allostasis proposes that long-term viability of a living system is achieved through anticipatory adjustments of its physiology and behaviour: emphasising physiological and affective stress as an adaptive state of adaptation that minimizes long-term prediction errors. More recently, the active inference framework (AIF) has also sought to explain action and long-term adaptation through the minimization of future errors (free energy), through the learning of statistical contingencies of the world, offering a formalism for allostatic regulation. We suggest that framing prediction errors through the lens of biological hormonal dynamics proposed by allostasis offers a way to integrate these two models together in a biologically-plausible manner. In this paper, we describe our initial work in developing a model that grounds prediction errors (surprisal) into the secretion of a physiological stress hormone (cortisol) acting as an adaptive, allostatic mediator on a homeostatically-controlled physiology. We evaluate this using a computational model in simulations using an active inference agent endowed with an artificial physiology, regulated through homeostatic and allostatic control in a stochastic environment. Our results find that allostatic functions of cortisol (stress), secreted as a function of prediction errors, provide adaptive advantages to the agent's long-term physiological regulation. We argue that the coupling of information-theoretic prediction errors to low-level, biological hormonal dynamics of stress can provide a computationally efficient model to long-term regulation for embodied intelligent systems., Comment: 14 pages, 4 figures

Can pretrained models generalize to new datasets without any retraining? We deploy pretrained image models on datasets they were not trained for, and investigate whether their embeddings form meaningful clusters. Our suite of benchmarking experiments use encoders pretrained solely on ImageNet-1k with either supervised or self-supervised training techniques, deployed on image datasets that were not seen during training, and clustered with conventional clustering algorithms. This evaluation provides new insights into the embeddings of self-supervised models, which prioritize different features to supervised models. Supervised encoders typically offer more utility than SSL encoders within the training domain, and vice-versa far outside of it, however, fine-tuned encoders demonstrate the opposite trend. Clustering provides a way to evaluate the utility of self-supervised learned representations orthogonal to existing methods such as kNN. Additionally, we find the silhouette score when measured in a UMAP-reduced space is highly correlated with clustering performance, and can therefore be used as a proxy for clustering performance on data with no ground truth labels. Our code implementation is available at \url{https://github.com/scottclowe/zs-ssl-clustering/}.

As the nation transitions from responding to the height of the COVID-19 pandemic to navigating its ongoing impact on youth and their communities, K-12 students continue to experience mental health and wellbeing challenges. In the fall of 2021, Redwood City School District (RCSD) placed a full-time, district-employed mental health counselor in each of its twelve schools in order to bolster district capacity to support mental health and wellbeing. This study aims to identify strategic opportunities for understanding and advancing the counseling program's impact as it moves into its third year. With these goals in mind, this research brief will describe findings related to three questions: (1) What is the role of RCSD's school-based mental health counselors?; (2) What are early signals or "lead indicators" of the value or impact of the counseling program?; and (3) What considerations might further improve the program as it moves into Year 3? To answer these questions, the authors conducted 19 interviews with district administrators, site administrators, and mental health counselors; and 22 participant observations of meetings related to the counseling program and student services more broadly. While the research team employed qualitative data analysis software (Dedoose 9.0.90) to facilitate the coding of qualitative data, the analysis of coded data followed a more traditional, iterative approach, moving between deductive and inductive reasoning. This research brief summarizes key findings and offers considerations for leveraging the counseling program to strengthen a district-wide, coordinated system of support for mental health and wellbeing and developing a theory-based evaluation of impact during Year 3 of program implementation. [This work is generously funded by Stanford's Office of the Provost through the Office of Community Engagement.]

To provide a landscape of mathematics activities children experience in U.S. preschool and kindergarten classrooms, we observed time children spent in mathematics activities (and -- as a contrast -- literacy) in 101 geographically diverse early childhood classrooms in seven U.S. states. We also observed what mathematics content, grouping strategies, and management formats teachers engaged during classroom mathematics activities. Each observation lasted approximately 2 hours; collectively these observations focused on 930 children observed one to three times during the 2018/2019 or 2019/2020 school year. Averaging across individuals' data within classrooms, we found that mathematics and literacy activities comprised 5% and 45% of time observed in preschool classrooms, respectively; and 26% and 42% of time observed in kindergarten classrooms, respectively. At both grades, when mathematics activities occurred, they were proportionally more often teacher-led rather than child-led. These findings raise concerns about the paucity of mathematics and over-reliance on developmentally inappropriate teacher-managed mathematics instruction in early childhood classrooms, especially preschools. Amount of time in math did not vary by preschool auspice, but time in literacy and the proportion of math time devoted to specific math content did: In Head Start classrooms we observed lowest percentage of time in literacy (and the highest percentage of time in non-instruction) compared to all other auspices. Across auspices, numeracy was the predominant math content area overall, but especially in Head Start classrooms. Thus, some aspects of early mathematics may differ with program auspice, suggesting that recommendations to increase and improve early mathematics activities may need to consider auspice characteristics. [This paper was published in "Early Childhood Research Quarterly" v69 p25-37 2024.]