Search

Cosmological information is encoded in the structure of galaxy clusters. In Universes with less matter and larger initial density perturbations, clusters form earlier and have more time to accrete material, leading to a more extended infall region. Thus, measuring the mean mass distribution in the infall region provides a novel cosmological test. The infall region is largely insensitive to baryonic physics, and provides a cleaner structural test than other measures of cluster assembly time such as concentration. We consider cluster samples from three publicly available galaxy cluster catalogues: the Spectrsopic Identification of eROSITA Sources (SPIDERS) catalogue, the X-ray and Sunyaev-Zeldovich effect selected clusters in the meta-catalogue M2C, and clusters identified in the Dark Energy Spectroscopic Instrument (DESI) Legacy Imaging Survey. Using a preliminary shape catalogue from the Ultraviolet Near Infrared Optical Northern Survey (UNIONS), we derive excess surface mass density profiles for each sample. We then compare the mean profile for the DESI Legacy sample, which is the most complete, to predictions from a suite of simulations covering a range of $\Omega_{\rm m}$ and $\sigma_8$, obtaining constraints of $\Omega_{\rm m}=0.29\pm 0.05$ and $\sigma_8=0.80 \pm 0.04$. We also measure mean (comoving) splashback radii for SPIDERS, M2C and DESI Legacy Imaging Survey clusters of $1.59^{+0.16}_{-0.13} {\rm cMpc}/h$, $1.30^{+0.25}_{-0.13} {\rm cMpc}/h$ and $1.45\pm0.11 {\rm cMpc}/h$ respectively. Performing this analysis with the final UNIONS shape catalogue and the full sample of spectroscopically observed clusters in DESI, we can expect to improve on the best current constraints from cluster abundance studies by a factor of 2 or more., Comment: 16 pages, 10 figures. Submitted to MNRAS

The theory of tensor categories has found applications across various fields, including representation theory, quantum field theory (conformal in 2 dimensions, and topological in 3 and 4 dimensions), quantum invariants of low-dimensional objects, topological phases of matter, and topological quantum computation. In essence, it is a categorification of the classical theory of algebras and modules. In this analogy, the Deligne tensor product $\boxtimes$ is to the linear tensor $\otimes_{\mathbb{C}}$ as the balanced tensor product $\boxtimes_C$ is to the tensor over algebra $\otimes_A$, where $\mathbb{C}$ is a field, $A$ is a $\mathbb{C}$-algebra, and $C$ is a tensor category. Before this work, several algebraic constructions for balanced tensor products were known, including categories of modules, internal Hom spaces, and generalized categorical centers. In this paper, we introduce a topological construction based on skein theory that offers a better mix of algebra and topology. This approach not only works for products of multiple module categories, but also provides the missing key to proving that the Turaev-Viro state sum model naturally arises from the 3-functor in the classification of fully extended field theories. Building on this result, we establish this long-anticipated proof in an upcoming work., Comment: 25 pages

Large Language Models, despite their significant capabilities, are known to fail in surprising and unpredictable ways. Evaluating their true `understanding' of language is particularly challenging due to the extensive web-scale data they are trained on. Therefore, we construct an evaluation to systematically assess natural language understanding (NLU) in LLMs by leveraging Construction Grammar (CxG), which provides insights into the meaning captured by linguistic elements known as constructions (Cxns). CxG is well-suited for this purpose because provides a theoretical basis to construct targeted evaluation sets. These datasets are carefully constructed to include examples which are unlikely to appear in pre-training data, yet intuitive and easy for humans to understand, enabling a more targeted and reliable assessment. Our experiments focus on downstream natural language inference and reasoning tasks by comparing LLMs' understanding of the underlying meanings communicated through 8 unique Cxns with that of humans. The results show that while LLMs demonstrate some knowledge of constructional information, even the latest models including GPT-o1 struggle with abstract meanings conveyed by these Cxns, as demonstrated in cases where test sentences are dissimilar to their pre-training data. We argue that such cases provide a more accurate test of true language understanding, highlighting key limitations in LLMs' semantic capabilities. We make our novel dataset and associated experimental data including prompts and model responses publicly available.

In a decision-making scenario, a principal could use conditional predictions from an expert agent to inform their choice. However, this approach would introduce a fundamental conflict of interest. An agent optimizing for predictive accuracy is incentivized to manipulate their principal towards more predictable actions, which prevents that principal from being able to deterministically select their true preference. We demonstrate that this impossibility result can be overcome through the joint evaluation of multiple agents. When agents are made to engage in zero-sum competition, their incentive to influence the action taken is eliminated, and the principal can identify and take the action they most prefer. We further prove that this zero-sum setup is unique, efficiently implementable, and applicable under stochastic choice. Experiments in a toy environment demonstrate that training on a zero-sum objective significantly enhances both predictive accuracy and principal utility, and can eliminate previously learned manipulative behavior.

We give an application of a Huisken monotonicity-type formula for the mean curvature flow in a compact smooth manifold with a Riemannian metric that evolves by a shrinking self-similar solution of the extended Ricci flow. Our investigation builds on previous articles by Huisken and the third author as we apply their techniques to establish new results in this geometric setting. Moreover, under some natural geometric assumptions, the noncompact case is also resolved, Comment: 17 pages

The o1 model series is trained with large-scale reinforcement learning to reason using chain of thought. These advanced reasoning capabilities provide new avenues for improving the safety and robustness of our models. In particular, our models can reason about our safety policies in context when responding to potentially unsafe prompts, through deliberative alignment. This leads to state-of-the-art performance on certain benchmarks for risks such as generating illicit advice, choosing stereotyped responses, and succumbing to known jailbreaks. Training models to incorporate a chain of thought before answering has the potential to unlock substantial benefits, while also increasing potential risks that stem from heightened intelligence. Our results underscore the need for building robust alignment methods, extensively stress-testing their efficacy, and maintaining meticulous risk management protocols. This report outlines the safety work carried out for the OpenAI o1 and OpenAI o1-mini models, including safety evaluations, external red teaming, and Preparedness Framework evaluations.

Weak gravitational lensing requires precise measurements of galaxy shapes and therefore an accurate knowledge of the PSF model. The latter can be a source of systematics that affect the shear two-point correlation function. A key stake of weak lensing analysis is to forecast the systematics due to the PSF. Correlation functions of galaxies and the PSF, the so-called $\rho$- and $\tau$-statistics, are used to evaluate the level of systematics coming from the PSF model and PSF corrections, and contributing to the two-point correlation function used to perform cosmological inference. Our goal is to introduce a fast and simple method to estimate this level of systematics and assess its agreement with state-of-the-art approaches. We introduce a new way to estimate the covariance matrix of the $\tau$-statistics using analytical expressions. The covariance allows us to estimate parameters directly related to the level of systematics associated with the PSF and provides us with a tool to validate the PSF model used in a weak-lensing analysis. We apply those methods to data from the Ultraviolet Near-Infrared Optical Northern Survey (UNIONS). We show that the semi-analytical covariance yields comparable results than using covariances obtained from simulations or jackknife resampling. It requires less computation time and is therefore well suited for rapid comparison of the systematic level obtained from different catalogs. We also show how one can break degeneracies between parameters with a redefinition of the $\tau$-statistics. The methods developed in this work will be useful tools in the analysis of current weak-lensing data but also of Stage IV surveys such as Euclid, LSST or Roman. They provide fast and accurate diagnostics on PSF systematics that are crucial to understand in the context of cosmic shear studies., Comment: 18 pages, 13 figures, submitted to A&A

Scaling has not yet been convincingly demonstrated for pure self-supervised learning from video. However, prior work has focused evaluations on semantic-related tasks $\unicode{x2013}$ action classification, ImageNet classification, etc. In this paper we focus on evaluating self-supervised learning on non-semantic vision tasks that are more spatial (3D) and temporal (+1D = 4D), such as camera pose estimation, point and object tracking, and depth estimation. We show that by learning from very large video datasets, masked auto-encoding (MAE) with transformer video models actually scales, consistently improving performance on these 4D tasks, as model size increases from 20M all the way to the largest by far reported self-supervised video model $\unicode{x2013}$ 22B parameters. Rigorous apples-to-apples comparison with many recent image and video models demonstrates the benefits of scaling 4D representations.

We introduce a hierarchical probabilistic approach to go from a 2D image to multiview 3D: a diffusion "prior" models the unseen 3D geometry, which then conditions a diffusion "decoder" to generate novel views of the subject. We use a pointmap-based geometric representation in a multiview image format to coordinate the generation of multiple target views simultaneously. We facilitate correspondence between views by assuming fixed target camera poses relative to the source camera, and constructing a predictable distribution of geometric features per target. Our modular, geometry-driven approach to novel-view synthesis (called "unPIC") beats SoTA baselines such as CAT3D and One-2-3-45 on held-out objects from ObjaverseXL, as well as real-world objects ranging from Google Scanned Objects, Amazon Berkeley Objects, to the Digital Twin Catalog.

In randomized trials and observational studies, it is often necessary to evaluate the extent to which an intervention affects a time-to-event outcome, which is only partially observed due to right censoring. For instance, in infectious disease studies, it is frequently of interest to characterize the relationship between risk of acquisition of infection with a pathogen and a biomarker previously measuring for an immune response against that pathogen induced by prior infection and/or vaccination. It is common to conduct inference within a causal framework, wherein we desire to make inferences about the counterfactual probability of survival through a given time point, at any given exposure level. To determine whether a causal effect is present, one can assess if this quantity differs by exposure level. Recent work shows that, under typical causal assumptions, summaries of the counterfactual survival distribution are identifiable. Moreover, when the treatment is multi-level, these summaries are also pathwise differentiable in a nonparametric probability model, making it possible to construct estimators thereof that are unbiased and approximately normal. In cases where the treatment is continuous, the target estimand is no longer pathwise differentiable, rendering it difficult to construct well-behaved estimators without strong parametric assumptions. In this work, we extend beyond the traditional setting with multilevel interventions to develop approaches to nonparametric inference with a continuous exposure. We introduce methods for testing whether the counterfactual probability of survival time by a given time-point remains constant across the range of the continuous exposure levels. The performance of our proposed methods is evaluated via numerical studies, and we apply our method to data from a recent pair of efficacy trials of an HIV monoclonal antibody., Comment: 22 pages, 2 figures

The population dynamics of the 229Th isomeric state is studied in a solid-state host under laser illumination. A photoquenching process is observed, where off-resonant vacuum-ultraviolet (VUV) radiation leads to relaxation of the isomeric state. The cross-section for this photoquenching process is measured and a model for the decay process, where photoexcitation of electronic states within the material bandgap opens an internal conversion decay channel, is presented and appears to reproduce the measured cross-section., Comment: 7 pages, 6 figures

In this work, we develop a simulation-based model to predict the excess surface mass density (ESD) depending on the local density environment. Using a conditional stellar mass function, our foreground galaxies are tailored toward the bright galaxy sample of the early data release of the Dark Energy Spectroscopic Instrument (DESI). Due to the nature of the ESD measurement, our derived model is directly applicable to all DESI data. To build this model, we use the $\texttt{AbacusSummit}$ N-body simulation suite from which we measure all necessary statistics and train an emulator based on $\texttt{CosmoPower}$. Finally, we present a cosmological parameter forecast for a possible combined analysis of DESI and the Ultraviolet Near Infrared Optical Northern Survey., Comment: 18 Pages, 16 Figures, submitted to Phys. Rev. D

While problem solving is a standard evaluation task for foundation models, a crucial component of problem solving -- actively and strategically gathering information to test hypotheses -- has not been closely investigated. To assess the information gathering abilities of foundation models in interactive environments, we introduce a framework in which a model must determine the factors influencing a hidden reward function by iteratively reasoning about its previously gathered information and proposing its next exploratory action to maximize information gain at each step. We implement this framework in both a text-based environment, which offers a tightly controlled setting and enables high-throughput parameter sweeps, and in an embodied 3D environment, which requires addressing complexities of multi-modal interaction more relevant to real-world applications. We further investigate whether approaches such as self-correction and increased inference time improve information gathering efficiency. In a relatively simple task that requires identifying a single rewarding feature, we find that LLM's information gathering capability is close to optimal. However, when the model must identify a conjunction of rewarding features, performance is suboptimal. The hit in performance is due partly to the model translating task description to a policy and partly to the model's effectiveness in using its in-context memory. Performance is comparable in both text and 3D embodied environments, although imperfect visual object recognition reduces its accuracy in drawing conclusions from gathered information in the 3D embodied case. For single-feature-based rewards, we find that smaller models curiously perform better; for conjunction-based rewards, incorporating self correction into the model improves performance.

We investigate the solutions of the relaxed MHD model (RxMHD) of Dewar \& Qu [J. Plasma Phys. {\bf 88}, 835880101 (2022)]. This model is a generalization of Taylor relaxation that allows the ideal Ohm's law constraint to be included, and this offers a pathway to extend the multi-region relaxed MHD (MRxMHD) model. By constructing solutions numerically, we show that the RxMHD model of Dewar \& Qu is mathematically well-defined and computationally feasible for constructing MHD equilibria. We also show that a cross-field flow can exist without enforcing an arbitrary constraint on the angular momentum (as is done in the case of MRxMHD with flow), and a pressure profile with a small gradient due to the Bernoulli flow. Our results also demonstrate the self-organization of fully relaxed regions during the optimization, which was an important motivation behind developing this model.

During their formation, galaxies are subject to tidal forces, which create correlations between their shapes and the large-scale structure of the Universe, known as intrinsic alignment. This alignment is a contamination for cosmic-shear measurements as one needs to disentangle correlations induced by external lensing effects from those intrinsically present in galaxies. We constrain the amplitude of intrinsic alignment and test models by making use of the overlap between the Ultraviolet Near-Infrared Optical Northern Survey (UNIONS) covering $3500 \, \mathrm{deg}^2$, and spectroscopic data from the Baryon Oscillation Spectroscopic Survey (BOSS/eBOSS). By comparing our results to measurements from other lensing surveys on the same spectroscopic tracers, we can test the reliability of these estimates and verify they are not survey dependent. We measure projected correlation functions between positions and ellipticities, which we model with perturbation theory to constrain the commonly used non-linear alignment model and its higher-order expansion. Using the non-linear alignment model, we obtain a $13\sigma$ detection with CMASS galaxies, a $3\sigma$ detection with LRGs, and a detection compatible with the null hypothesis for ELGs. We test the tidal alignment and tidal torque model, a higher-order alignment model, which we find to be in good agreement with the non-linear alignment prediction and for which we can constrain the second-order parameters. We show a strong scaling of our intrinsic alignment amplitude with luminosity. We demonstrate that the UNIONS sample is robust against systematic contributions, particularly concerning PSF biases. We reached a reasonable agreement when comparing our measurements to other lensing samples for the same spectroscopic samples. We take this agreement as an indication that direct measurements of intrinsic alignment are mature for stage IV priors., Comment: To be submitted to A&A, 18 pages, 17 figures

We present Track Anything Behind Everything (TABE), a novel dataset, pipeline, and evaluation framework for zero-shot amodal completion from visible masks. Unlike existing methods that require pretrained class labels, our approach uses a single query mask from the first frame where the object is visible, enabling flexible, zero-shot inference. Our dataset, TABE-51 provides highly accurate ground truth amodal segmentation masks without the need for human estimation or 3D reconstruction. Our TABE pipeline is specifically designed to handle amodal completion, even in scenarios where objects are completely occluded. We also introduce a specialised evaluation framework that isolates amodal completion performance, free from the influence of traditional visual segmentation metrics.

Entanglement often increases quantum measurement schemes' sensitivity. However, we find that in precision measurements with zero-mean Gaussian states, such as squeezed states, entanglement between different paths degrades measurement sensitivity. We prove an inverse relationship between entanglement entropy and sensitivity for measurements of single-mode phase shifts in multimode systems and for phase shifts on both modes in two-mode systems. In the two-mode case, which models devices such as interferometers, we find that entanglement strongly degrades differential phase sensitivity. Finally, we show that minimizing entanglement between paths maximizes the phase sensitivity of $N$-mode systems with zero-mean Gaussian state inputs., Comment: 5+6 pages, 1 figure

Laser excitation of thorium-229 nuclei in doped wide bandgap crystals has been demonstrated recently, opening the possibility of developing ultrastable solid-state clocks and sensitive searches for new physics. We develop a quantitative theory of the internal conversion of isomeric thorium-229 in solid-state hosts. The internal conversion of the isomer proceeds by resonantly exciting a valence band electron to a defect state, accompanied by multi-phonon emission. We demonstrate that, if the process is energetically allowed, it generally quenches the isomer on timescales much faster than the isomer's radiative lifetime, despite thorium being in the +4 charge state in the valence band.

Recent advances in semiconductor spin qubits have achieved linear arrays exceeding ten qubits. Moving to two-dimensional (2D) qubit arrays is a critical next step to advance towards fault-tolerant implementations, but it poses substantial fabrication challenges, particularly because enabling control of nearest-neighbor entanglement requires the incorporation of interstitial exchange gates between quantum dots in the qubit architecture. In this work, we present a 2D array of silicon metal-oxide-semiconductor (MOS) quantum dots with tunable interdot coupling between all adjacent dots. The device is characterized at 4.2 K, where we demonstrate the formation and isolation of double-dot and triple-dot configurations. We show control of all nearest-neighbor tunnel couplings spanning up to 30 decades per volt through the interstitial exchange gates and use advanced modeling tools to estimate the exchange interactions that could be realized among qubits in this architecture. These results represent a significant step towards the development of 2D MOS quantum processors compatible with foundry manufacturing techniques., Comment: 9 pages, 5 figures

State-of-the-art algorithms are reported to be almost perfect at distinguishing the vibrations arising from healthy and damaged machine bearings, according to benchmark datasets at least. However, what about their application to new data? In this paper, we are able to confirm that neural networks for bearing fault detection can be crippled by incorrect hyperparameterisation, and also that the correct hyperparameter settings can actually change when transitioning to new data. The paper weaves together multiple methods to explain the behaviour of the hyperparameters of a wide-kernel convolutional neural network and how to set them. Since guidance already exists for generic hyperparameters like minibatch size, we focus on how to set architecture-specific hyperparameters such as the width of the convolutional kernels, a topic which might otherwise be obscure. We reflect different data properties by fusing information from seven different benchmark datasets, and our results show that the kernel size in the first layer in particular is sensitive to changes in the data. Looking deeper, we use manipulated copies of one dataset in an attempt to spot why the kernel size sometimes needs to change. The relevance of sampling rate is studied by using different levels of resampling, and spectral content is studied by increasingly filtering out high frequencies. At the end of our paper we conclude by stating clear guidance on how to set the hyperparameters of our neural network architecture., Comment: 71 pages, 14 figures, 7 tables

We introduce the Situated Corpus Of Understanding Transactions (SCOUT), a multi-modal collection of human-robot dialogue in the task domain of collaborative exploration. The corpus was constructed from multiple Wizard-of-Oz experiments where human participants gave verbal instructions to a remotely-located robot to move and gather information about its surroundings. SCOUT contains 89,056 utterances and 310,095 words from 278 dialogues averaging 320 utterances per dialogue. The dialogues are aligned with the multi-modal data streams available during the experiments: 5,785 images and 30 maps. The corpus has been annotated with Abstract Meaning Representation and Dialogue-AMR to identify the speaker's intent and meaning within an utterance, and with Transactional Units and Relations to track relationships between utterances to reveal patterns of the Dialogue Structure. We describe how the corpus and its annotations have been used to develop autonomous human-robot systems and enable research in open questions of how humans speak to robots. We release this corpus to accelerate progress in autonomous, situated, human-robot dialogue, especially in the context of navigation tasks where details about the environment need to be discovered., Comment: 14 pages, 7 figures

In this paper, we describe the development of symbolic representations annotated on human-robot dialogue data to make dimensions of meaning accessible to autonomous systems participating in collaborative, natural language dialogue, and to enable common ground with human partners. A particular challenge for establishing common ground arises in remote dialogue (occurring in disaster relief or search-and-rescue tasks), where a human and robot are engaged in a joint navigation and exploration task of an unfamiliar environment, but where the robot cannot immediately share high quality visual information due to limited communication constraints. Engaging in a dialogue provides an effective way to communicate, while on-demand or lower-quality visual information can be supplemented for establishing common ground. Within this paradigm, we capture propositional semantics and the illocutionary force of a single utterance within the dialogue through our Dialogue-AMR annotation, an augmentation of Abstract Meaning Representation. We then capture patterns in how different utterances within and across speaker floors relate to one another in our development of a multi-floor Dialogue Structure annotation schema. Finally, we begin to annotate and analyze the ways in which the visual modalities provide contextual information to the dialogue for overcoming disparities in the collaborators' understanding of the environment. We conclude by discussing the use-cases, architectures, and systems we have implemented from our annotations that enable physical robots to autonomously engage with humans in bi-directional dialogue and navigation., Comment: 52 pages, 14 figures

Alkali-metal-noble-gas comagnetometers are precision probes well-suited for tests of fundamental physics and inertial rotation sensing, combining high sensitivity of the spin-exchange-relaxation free (SERF) magnetometers with inherent suppression of magnetic field noise. Past versions of the device utilizing continuous-wave optical pumping are sensitive to a single axis perpendicular to the plane spanned by the orthogonal pump and probe laser beams. These devices are susceptible to light shifts in the alkali atoms, and to power and beam pointing fluctuations of both the probe and pump lasers, the latter of which is a dominant source of $1/f$ noise. In this work, we model and implement a new approach to alkali-metal-noble-gas comagnetometers using pulsed optical pumping. After each pump laser pulse, an off-resonance probe beam measures the precession of noble-gas-spins-coupled alkali spins via optical rotation in the dark, thus eliminating effects from pump laser light shift and power fluctuations. Performing non-linear fitting on the sinusoidal transient signal with a proper phase enables separate and simultaneous measurement of signals along two orthogonal axes in the plane perpendicular to the pump beam. Effects from beam pointing fluctuations of the probe beam in the pump-probe plane is fundamentally eliminated, and signal response to pump beam pointing fluctuations is suppressed by compensation from noble-gas nuclear spins., Comment: 12 pages, 11 figures

Multimodal learning, which involves integrating information from various modalities such as text, images, audio, and video, is pivotal for numerous complex tasks like visual question answering, cross-modal retrieval, and caption generation. Traditional approaches rely on modality-specific encoders and late fusion techniques, which can hinder scalability and flexibility when adapting to new tasks or modalities. To address these limitations, we introduce a novel framework that extends the concept of task reformulation beyond natural language processing (NLP) to multimodal learning. We propose to reformulate diverse multimodal tasks into a unified next-frame prediction problem, allowing a single model to handle different modalities without modality-specific components. This method treats all inputs and outputs as sequential frames in a video, enabling seamless integration of modalities and effective knowledge transfer across tasks. Our approach is evaluated on a range of tasks, including text-to-text, image-to-text, video-to-video, video-to-text, and audio-to-text, demonstrating the model's ability to generalize across modalities with minimal adaptation. We show that task reformulation can significantly simplify multimodal model design across various tasks, laying the groundwork for more generalized multimodal foundation models., Comment: 10 pages, 10 figures

Environmental and individualistic variables affect the rate of human decomposition in complex ways. These effects complicate the estimation of the postmortem interval (PMI) based on observed decomposition characteristics. In this work, we develop a generative probabilistic model for decomposing human remains based on PMI and a wide range of environmental and individualistic variables. This model explicitly represents the effect of each variable, including PMI, on the appearance of each decomposition characteristic, allowing for direct interpretation of model effects and enabling the use of the model for PMI inference and optimal experimental design. In addition, the probabilistic nature of the model allows for the integration of expert knowledge in the form of prior distributions. We fit this model to a diverse set of 2,529 cases from the GeoFOR dataset. We demonstrate that the model accurately predicts 24 decomposition characteristics with an ROC AUC score of 0.85. Using Bayesian inference techniques, we invert the decomposition model to predict PMI as a function of the observed decomposition characteristics and environmental and individualistic variables, producing an R-squared measure of 71%. Finally, we demonstrate how to use the fitted model to design future experiments that maximize the expected amount of new information about the mechanisms of decomposition using the Expected Information Gain formalism.

We calculate the ordinary $C_2$-cohomology of $BT^2$ with Burnside ring coefficients, using an extended grading that allows us to capture a more natural set of generators. We discuss how this cohomology is related to those of $BT^1$ and $BU(2)$, calculated previously, both relationships being more complicated than in the nonequivariant case., Comment: 45 pages

Current vision models typically maintain a fixed correspondence between their representation structure and image space. Each layer comprises a set of tokens arranged "on-the-grid," which biases patches or tokens to encode information at a specific spatio(-temporal) location. In this work we present Moving Off-the-Grid (MooG), a self-supervised video representation model that offers an alternative approach, allowing tokens to move "off-the-grid" to better enable them to represent scene elements consistently, even as they move across the image plane through time. By using a combination of cross-attention and positional embeddings we disentangle the representation structure and image structure. We find that a simple self-supervised objective--next frame prediction--trained on video data, results in a set of latent tokens which bind to specific scene structures and track them as they move. We demonstrate the usefulness of MooG's learned representation both qualitatively and quantitatively by training readouts on top of the learned representation on a variety of downstream tasks. We show that MooG can provide a strong foundation for different vision tasks when compared to "on-the-grid" baselines., Comment: Accepted to NeurIPS 2024 (spotlight). Project page: https://moog-paper.github.io/

Even if the magnetic field in a stellarator is integrable, phase-space integrability for energetic particle guiding center trajectories is not guaranteed. Both trapped and passing particle trajectories can experience convective losses, caused by wide phase-space island formation, and diffusive losses, caused by phase-space island overlap. By locating trajectories that are closed in the angle coordinate but not necessarily closed in the radial coordinate, we can quantify the magnitude of the perturbation that results in island formation. We characterize island width and island overlap in quasihelical (QH) and quasiaxisymmetric (QA) finite-beta equilibria for both trapped and passing energetic particles. For trapped particles in QH, low-shear toroidal precession frequency profiles near zero result in wide island formation. While QA transit frequencies do not cross through the zero resonance, we observe that island overlap is more likely since higher shear results in the crossing of more low-order resonances.

Gas-phase molecules capable of repeatable, narrow-band spontaneous photon scattering are prized for direct laser cooling and quantum state detection. Recently, large molecules incorporating phenyl rings have been shown to exhibit similar vibrational closure to the small molecules demonstrated so far, and it is not yet known if the high vibrational-mode density of even larger species will eventually compromise optical cycling. Here, we systematically increase the size of hydrocarbon ligands attached to single alkaline-earth-phenoxides from (-H) to -C$_{14}$H$_{19}$ while measuring the vibrational branching fractions of the optical transition. We find that varying the ligand size from 1 to more than 30 atoms does not systematically reduce the cycle closure, which remains around 90%. Theoretical extensions to larger diamondoids and bulk diamond surface suggest that alkaline earth phenoxides may maintain the desirable scattering behavior as the system size grows further, with no indication of an upper limit., Comment: 9 pages, 5 figures, SI(34 pages)

The recent laser excitation of the 229Th isomeric transition in a solid-state host opens the door for a portable solid-state nuclear optical clock. However, at present the vacuum-ultraviolet laser systems required for clock operation are not conducive to a fieldable form factor. Here, we propose a possible solution to this problem by using 229Th-doped nonlinear optical crystals, which would allow clock operation without a vacuum-ultraviolet laser system and without the need of maintaining the crystal under vacuum.

GPT-4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50\% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities.

Fabrication of quantum processors in advanced 300 mm wafer-scale complementary metal-oxide-semiconductor (CMOS) foundries provides a unique scaling pathway towards commercially viable quantum computing with potentially millions of qubits on a single chip. Here, we show precise qubit operation of a silicon two-qubit device made in a 300 mm semiconductor processing line. The key metrics including single- and two-qubit control fidelities exceed 99% and state preparation and measurement fidelity exceeds 99.9%, as evidenced by gate set tomography (GST). We report coherence and lifetimes up to $T_\mathrm{2}^{\mathrm{*}} = 30.4$ $\mu$s, $T_\mathrm{2}^{\mathrm{Hahn}} = 803$ $\mu$s, and $T_1 = 6.3$ s. Crucially, the dominant operational errors originate from residual nuclear spin carrying isotopes, solvable with further isotopic purification, rather than charge noise arising from the dielectric environment. Our results answer the longstanding question whether the favourable properties including high-fidelity operation and long coherence times can be preserved when transitioning from a tailored academic to an industrial semiconductor fabrication technology., Comment: 10 pages, 4 figures, 4 extended data figures

The Workflows Community Summit gathered 111 participants from 18 countries to discuss emerging trends and challenges in scientific workflows, focusing on six key areas: time-sensitive workflows, AI-HPC convergence, multi-facility workflows, heterogeneous HPC environments, user experience, and FAIR computational workflows. The integration of AI and exascale computing has revolutionized scientific workflows, enabling higher-fidelity models and complex, time-sensitive processes, while introducing challenges in managing heterogeneous environments and multi-facility data dependencies. The rise of large language models is driving computational demands to zettaflop scales, necessitating modular, adaptable systems and cloud-service models to optimize resource utilization and ensure reproducibility. Multi-facility workflows present challenges in data movement, curation, and overcoming institutional silos, while diverse hardware architectures require integrating workflow considerations into early system design and developing standardized resource management tools. The summit emphasized improving user experience in workflow systems and ensuring FAIR workflows to enhance collaboration and accelerate scientific discovery. Key recommendations include developing standardized metrics for time-sensitive workflows, creating frameworks for cloud-HPC integration, implementing distributed-by-design workflow modeling, establishing multi-facility authentication protocols, and accelerating AI integration in HPC workflow management. The summit also called for comprehensive workflow benchmarks, workflow-specific UX principles, and a FAIR workflow maturity model, highlighting the need for continued collaboration in addressing the complex challenges posed by the convergence of AI, HPC, and multi-facility research environments.

Understanding neural networks is crucial to creating reliable and trustworthy deep learning models. Most contemporary research in interpretability analyzes just one model at a time via causal intervention or activation analysis. Yet despite successes, these methods leave significant gaps in our understanding of the training behaviors of neural networks, how their inner representations emerge, and how we can predictably associate model components with task-specific behaviors. Seeking new insights from work in related fields, here we survey literature in the field of model merging, a field that aims to combine the abilities of various neural networks by merging their parameters and identifying task-specific model components in the process. We analyze the model merging literature through the lens of loss landscape geometry, an approach that enables us to connect observations from empirical studies on interpretability, security, model merging, and loss landscape analysis to phenomena that govern neural network training and the emergence of their inner representations. To systematize knowledge in this area, we present a novel taxonomy of model merging techniques organized by their core algorithmic principles. Additionally, we distill repeated empirical observations from the literature in these fields into characterizations of four major aspects of loss landscape geometry: mode convexity, determinism, directedness, and connectivity. We argue that by improving our understanding of the principles underlying model merging and loss landscape geometry, this work contributes to the goal of ensuring secure and trustworthy machine learning in practice.

Visual Simultaneous Localization and Mapping (VSLAM) is a fundamental technology for robotics applications. While VSLAM research has achieved significant advancements, its robustness under challenging situations, such as poor lighting, dynamic environments, motion blur, and sensor failures, remains a challenging issue. To address these challenges, we introduce a novel RGB-D dataset designed for evaluating the robustness of VSLAM systems. The dataset comprises real-world indoor scenes with dynamic objects, motion blur, and varying illumination, as well as emulated camera failures, including lens dirt, condensation, underexposure, and overexposure. Additionally, we offer open-source scripts for injecting camera failures into any images, enabling further customization by the research community. Our experiments demonstrate that ORB-SLAM2, a traditional VSLAM algorithm, and TartanVO, a Deep Learning-based VO algorithm, can experience performance degradation under these challenging conditions. Therefore, this dataset and the camera failure open-source tools provide a valuable resource for developing more robust VSLAM systems capable of handling real-world challenges., Comment: 6 pages

Linear time-translation-invariant (LTI) models offer simple, yet powerful, abstractions of complex classical dynamical systems. Quantum versions of such models have so far relied on assumptions of Markovianity or an internal state-space description. We develop a general quantization scheme for multimode classical LTI systems that reveals their fundamental quantum noise, is applicable to non-Markovian scenarios, and does not require knowledge of an internal description. The resulting model is that of an open quantum LTI system whose dilation to a closed system is characterized by elements of the conjugate symplectic group. Using Lie group techniques, we show that such systems can be synthesized using frequency-dependent interferometers and squeezers. We derive tighter Heisenberg uncertainty bounds, which constrain the ultimate performance of any LTI system, and obtain an invariant representation of their output noise covariance matrix that reveals the ubiquity of "complex squeezing" in lossy systems. This frequency-dependent quantum resource can be hidden to homodyne and heterodyne detection and can only be revealed with more general "symplectodyne" detection. These results establish a complete and systematic framework for the analysis, synthesis, and measurement of arbitrary quantum LTI systems., Comment: 14+6 pages, 6 figures. See ancillary file for Supplementary Information

Spin precession is a textbook example of dynamics of a quantum system that exactly mimics its classical counterpart. Here we challenge this view by certifying the quantumness of exotic states of a nuclear spin through its uniform precession. The key to this result is measuring the positivity, instead of the expectation value, of the $x$-projection of the precessing spin, and using a spin > 1/2 qudit, that is not restricted to semi-classical spin coherent states. The experiment is performed on a single spin-7/2 $^{123}$Sb nucleus, implanted in a silicon nanoelectronic device, amenable to high-fidelity preparation, control, and projective single-shot readout. Using Schr\"odinger cat states and other bespoke states of the nucleus, we violate the classical bound by 19 standard deviations, proving that no classical probability distribution can explain the statistic of this spin precession, and highlighting our ability to prepare quantum resource states with high fidelity in a single atomic-scale qudit., Comment: Main text: 11 pages, 5 figures. Supplementary information: 13 pages, 11 figures

After nearly fifty years of searching, the vacuum ultraviolet $^{229}$Th nuclear isomeric transition has recently been directly laser excited [1,2] and measured with high spectroscopic precision [3]. Nuclear clocks based on this transition are expected to be more robust [4,5] than and may outperform [6,7] current optical atomic clocks. They also promise sensitive tests for new physics beyond the standard model [5,8,9]. In light of these important advances and applications, a dramatic increase in the need for $^{229}$Th spectroscopy targets in a variety of platforms is anticipated. However, the growth and handling of high-concentration $^{229}$Th-doped crystals [5] used in previous measurements [1-3,10] are challenging due to the scarcity and radioactivity of the $^{229}$Th material. Here, we demonstrate a potentially scalable solution to these problems by demonstrating laser excitation of the nuclear transition in $^{229}$ThF$_4$ thin films grown with a physical vapor deposition process, consuming only micrograms of $^{229}$Th material. The $^{229}$ThF$_4$ thin films are intrinsically compatible with photonics platforms and nanofabrication tools for integration with laser sources and detectors, paving the way for an integrated and field-deployable solid-state nuclear clock with radioactivity up to three orders of magnitude smaller than typical \thor-doped crystals [1-3,10]. The high nuclear emitter density in $^{229}$ThF$_4$ also potentially enables quantum optics studies in a new regime. Finally, we describe the operation and present the estimation of the performance of a nuclear clock based on a defect-free ThF$_4$ crystal., Comment: 15 pages, 3 figures

Objectives: To identify a head acceleration event (HAE) severity measure associated with HIA1 removals in elite level rugby union. Methods: HAEs were recorded from 215 men and 325 women with 30 and 28 HIA1 removals from men and women, respectively. Logistical regression were calculated to identify if peak power, maximum principal strain (MPS) and or Head Acceleration Response Metric (HARM) were associated with HIA1 events compared to non-cases. Optimal threshold values were determined using the Youden Index. Area under the curve (AUC) were compared using a paired sample approach. Significant differences were set at p<0.05. Results: All three severity measures were associated with HIA1 removals in both the mens and womens game. Power performed greatest for HIA1 removals in both the mens and womens games, based on overall AUC, sensitivity, and specificity values. HARM and MPS were found to perform lower than PLA in the womens game based on AUC comparisons (p=0.006 and 0.001, respectively), with MPS performing lower than PAA (p=0.001). Conclusion: The findings progress our understanding of HAE measures associated with HIA1 removals. Peak power, a measure based on fundamental mechanics and commonly used in sports performance, may be a suitable HAE severity measure., Comment: 4 Tables, 2 Figures

Language models (LMs) can "memorize" information, i.e., encode training data in their weights in such a way that inference-time queries can lead to verbatim regurgitation of that data. This ability to extract training data can be problematic, for example, when data are private or sensitive. In this work, we investigate methods to mitigate memorization: three regularizer-based, three finetuning-based, and eleven machine unlearning-based methods, with five of the latter being new methods that we introduce. We also introduce TinyMem, a suite of small, computationally-efficient LMs for the rapid development and evaluation of memorization-mitigation methods. We demonstrate that the mitigation methods that we develop using TinyMem can successfully be applied to production-grade LMs, and we determine via experiment that: regularizer-based mitigation methods are slow and ineffective at curbing memorization; fine-tuning-based methods are effective at curbing memorization, but overly expensive, especially for retaining higher accuracies; and unlearning-based methods are faster and more effective, allowing for the precise localization and removal of memorized information from LM weights prior to inference. We show, in particular, that our proposed unlearning method BalancedSubnet outperforms other mitigation methods at removing memorized information while preserving performance on target tasks.

Recent work has focused on nonparametric estimation of conditional treatment effects, but inference has remained relatively unexplored. We propose a class of nonparametric tests for both quantitative and qualitative treatment effect heterogeneity. The tests can incorporate a variety of structured assumptions on the conditional average treatment effect, allow for both continuous and discrete covariates, and do not require sample splitting. Furthermore, we show how the tests are tailored to detect alternatives where the population impact of adopting a personalized decision rule differs from using a rule that discards covariates. The proposal is thus relevant for guiding treatment policies. The utility of the proposal is borne out in simulation studies and a re-analysis of an AIDS clinical trial.

This work introduces a quantitative model for the Frank-Read source, considered to be one of the most important micro-mechanical mechanisms of dislocation creation in crystalline materials. It has long been known that these sources create dislocations in a repetitive, oscillatory process, which is driven by an external shear force. Unlike the existing explanations in the literature, the model introduced in the present article is based on just a few simple physical principles, namely line tension and dislocation motion due to a single slip plane flow rule, together with a pinning constraint on the ends of the central dislocation line. A complete discretisation, including suitable re-meshing and 'topological cutting' algorithms, is described and simulation results are discussed. Despite its conceptual simplicity, the model and discretisation described in the present work yield remarkably accurate predictions about the shape and properties of the Frank-Read source. In particular, it is shown that only one dimensionless parameter controls the dynamics of the Frank-Read source if one neglects crystal anisotropy. This allows to derive an emergent law about the length of dislocation line generated per shear energy., Comment: 20 pages, 8 figures

Significant work has been done to develop quantum satellites, which generate entangled pairs in space and distribute them to ground stations separated some distance away. The reverse uplink case, where pairs are generated on the ground and swapped on the satellite using an optical Bell-measurement, has not been seriously considered due to a prevailing assumption that it is practically infeasible. In this letter, we illustrate the feasibility of performing Discrete Variable photonic Bell-measurements in space by conducting a detailed numerical analysis to estimate the channel efficiency and attainable pair fidelity for various satellite-station configurations. Our model accounts for a wide range of physical effects such as atmospheric effects, stray photons, and mode mismatch. Our findings show promise toward the feasibility of photonic Bell-measurements in space, which motivates future research towards large-scale Satellite-based uplink entanglement distribution., Comment: 6 pages, 3 figures

Federated Learning (FL) is a decentralized machine learning paradigm where models are trained on distributed devices and are aggregated at a central server. Existing FL frameworks assume simple two-tier network topologies where end devices are directly connected to the aggregation server. While this is a practical mental model, it does not exploit the inherent topology of real-world distributed systems like the Internet-of-Things. We present Flight, a novel FL framework that supports complex hierarchical multi-tier topologies, asynchronous aggregation, and decouples the control plane from the data plane. We compare the performance of Flight against Flower, a state-of-the-art FL framework. Our results show that Flight scales beyond Flower, supporting up to 2048 simultaneous devices, and reduces FL makespan across several models. Finally, we show that Flight's hierarchical FL model can reduce communication overheads by more than 60%.

As large language models (LLMs) gain traction in healthcare, concerns about their susceptibility to demographic biases are growing. We introduce {DiversityMedQA}, a novel benchmark designed to assess LLM responses to medical queries across diverse patient demographics, such as gender and ethnicity. By perturbing questions from the MedQA dataset, which comprises medical board exam questions, we created a benchmark that captures the nuanced differences in medical diagnosis across varying patient profiles. Our findings reveal notable discrepancies in model performance when tested against these demographic variations. Furthermore, to ensure the perturbations were accurate, we also propose a filtering strategy that validates each perturbation. By releasing DiversityMedQA, we provide a resource for evaluating and mitigating demographic bias in LLM medical diagnoses., Comment: Published in NLP4PI @ EMNLP 2024, Accepted to AIM-FM @ NeurIPS 2024

Being three-dimensional, stellarators have the advantage that plasma currents are not essential for creating rotational-transform; however, the external current-carrying coils in stellarators can have strong geometrical shaping, which can complicate the construction. Reducing the inter-coil electromagnetic forces acting on strongly shaped 3D coils and the stress on the support structure, while preserving the favorable properties of the magnetic field is a design challenge. In this work, we recognize that the inter-coil ${\boldsymbol{j}} \times {\boldsymbol{B}}$ forces are the gradient of the vacuum magnetic energy, $\displaystyle E := \frac{1}{2\mu_0}\int_{\mathbb{R}^3} \!\!\! B^2 \, dV$. We introduce an objective functional, ${\mathcal{F}}:= \Phi_2 + \omega E$, built on the usual quadratic flux on a prescribed target surface, $\displaystyle \Phi_2 := \frac{1}{2}\int_{\mathcal{S}} ( {\boldsymbol{B}} \cdot {\boldsymbol{n}} )^2 \, dS$, and the vacuum energy, where $\omega$ is a weight penalty. The Euler-Lagrange equation for stationary states is derived, and numerical illustrations are computed using the SIMSOPT code \cite{simsopt}.

We report a systematic comparison of niobium nitride thin films deposited on oxidized silicon substrates by reactive DC magnetron sputtering and reactive high-power impulse magnetron sputtering (HiPIMS). After determining the nitrogen gas concentration that produces the highest superconducting critical temperature for each process, we characterize the dependence of the critical temperature on film thickness. The optimal nitrogen concentration is higher for HiPIMS than for DC sputtering, and HiPIMS produces higher critical temperatures for all thicknesses studied. We attribute this to the HiPIMS process enabling the films to get closer to optimal stoichiometry before beginning to form a hexagonal crystal phase that reduces the critical temperature, along with the extra kinetic energy in the HiPIMS process improving crystallinity. We also study the ability to increase the critical temperature of the HiPIMS films through the use of an aluminum nitride buffer layer and substrate heating., Comment: 7 pages, 7 figures