Search

In order to understand the effects of a space radiation environment on cross-strip germanium detectors, we investigated the effects of high-energy proton damage on a COSI detector and the capabilities of high-temperature annealing in repairing detector spectral resolution. We irradiated a COSI-balloon cross-strip high-purity germanium (HPGe) detector with a high-energy proton fluence corresponding to ~10 years in a space radiation environment. We repaired the resulting degradation in spectral resolution within 16% of its preradiation value through a series of high-temperature anneals. We characterize the repair of charge traps with time spent under high-temperature anneal to inform an annealing procedure for long-term maintenance of COSI's spectral resolution., Comment: 22 pages, 7 figures, 2 tables, to be published in NIM A

In doped transition metal dichalcogenides, optically created excitons (bound electron-hole pairs) can strongly interact with a Fermi sea of electrons to form Fermi polaron quasiparticles. When there are two distinct Fermi seas, as is the case in WSe$_2$, there are two flavors of lowest-energy (attractive) polarons -- singlet and triplet -- where the exciton is coupled to the Fermi sea in the same or opposite valley, respectively. Using two-dimensional coherent electronic spectroscopy, we analyze how their quantum decoherence evolves with doping density and determine the condition under which stable Fermi polarons form. Because of the large oscillator strength associated with these resonances, intrinsic quantum dynamics of polarons as well as valley coherence between coupled singlet- and triplet polarons occur on sub-picosecond time scales. Surprisingly, we find that a dark-to-bright state conversion process leads to a particularly long-lived singlet polaron valley polarization, persisting up to 200-800 ps. Valley coherence between the singlet- and triplet polaron is correlated with their energy fluctuations. Our finding provides valuable guidance for the electrical and optical control of spin and valley indexes in atomically thin semiconductors., Comment: 4 pages, 4 figures

Background: Late Gadolinium Enhancement (LGE) imaging is the gold standard for assessing myocardial fibrosis and scarring, with left ventricular (LV) LGE extent predicting major adverse cardiac events (MACE). Despite its importance, routine LGE-based LV scar quantification is hindered by labor-intensive manual segmentation and inter-observer variability. Methods: We propose ScarNet, a hybrid model combining a transformer-based encoder from the Medical Segment Anything Model (MedSAM) with a convolution-based U-Net decoder, enhanced by tailored attention blocks. ScarNet was trained on 552 ischemic cardiomyopathy patients with expert segmentations of myocardial and scar boundaries and tested on 184 separate patients. Results: ScarNet achieved robust scar segmentation in 184 test patients, yielding a median Dice score of 0.912 (IQR: 0.863--0.944), significantly outperforming MedSAM (median Dice = 0.046, IQR: 0.043--0.047) and nnU-Net (median Dice = 0.638, IQR: 0.604--0.661). ScarNet demonstrated lower bias (-0.63%) and coefficient of variation (4.3%) compared to MedSAM (bias: -13.31%, CoV: 130.3%) and nnU-Net (bias: -2.46%, CoV: 20.3%). In Monte Carlo simulations with noise perturbations, ScarNet achieved significantly higher scar Dice (0.892 \pm 0.053, CoV = 5.9%) than MedSAM (0.048 \pm 0.112, CoV = 233.3%) and nnU-Net (0.615 \pm 0.537, CoV = 28.7%). Conclusion: ScarNet outperformed MedSAM and nnU-Net in accurately segmenting myocardial and scar boundaries in LGE images. The model exhibited robust performance across diverse image qualities and scar patterns., Comment: 31 pages, 8 figures

The Adomian decomposition method (ADM) is a universal approach to solving governing equations in various engineering and technological applications. The applicability of the ADM is almost limitless due to its universal applicability, but its convergence rate and numerical accuracy are sensitive to the number of truncated terms in series solutions. More importantly, Adomian formalism still holds unresolved issues regarding the mismatch of the order of the expansion parameter. The current work provides an in-depth analysis of Adomian's decomposition method, Lyapunov's stability theory, and the nonlinear perturbation theory to resolve the fundamental mismatch with physical interpretation.

Exceptional points of a dissipative chain of three coupled oscillators (trimer), which is driven by quadratic photon, are investigated. The exceptional points emerge from the coalescence of both eigenvalues and eigenvectors of the dynamical matrix that describes the first moments of the trimer. At the exceptional point, we found that the optical spectrum is split into two peaks, instead of a conventional single peak, as in the case of a single oscillator. In particular, the positions of these peaks correspond to the natural frequency of the trimer in a \textit{closed system}, which depends only on the coupling strength. Furthermore, after passing the exceptional point, the peak positions do not change, which can be used to estimate the coupling strength between oscillators.

We present a study of very low-mass stars and brown dwarfs in the rich star-forming core of the Rho Ophiuchi cloud complex. The selection of the sample relies on detecting the inherent water absorption characteristic in young substellar objects. Of the 22 water-bearing candidates selected, 15 have a spectral type of M6 or later. Brown dwarf candidates too faint for membership determination by Gaia have their proper motions derived by deep-infrared images spanning six years. Astrometric analysis confirms 21/22 sources as members, one identified as a contaminant. Infrared colors and the spectral energy distribution of each water-bearing candidate are used to diagnose the mass, age, and possible existence of circumstellar dust. 15 sources exhibit evidence of disks in their spectral energy distributions, as late as in M8-type objects. Spectroscopy for bright candidates has confirmed one as an M8 member and verified two sources (with disks) exhibiting signatures of magnetospheric accretion., Comment: 25 pages, 35 figures, 4 tables. Accepted for publication in The Astrophysical Journal

Artificial intelligence (AI) has enabled agents to master complex video games, from first-person shooters like Counter-Strike to real-time strategy games such as StarCraft II and racing games like Gran Turismo. While these achievements are notable, applying these AI methods in commercial video game production remains challenging due to computational constraints. In commercial scenarios, the majority of computational resources are allocated to 3D rendering, leaving limited capacity for AI methods, which often demand high computational power, particularly those relying on pixel-based sensors. Moreover, the gaming industry prioritizes creating human-like behavior in AI agents to enhance player experience, unlike academic models that focus on maximizing game performance. This paper introduces a novel methodology for training neural networks via imitation learning to play a complex, commercial-standard, VALORANT-like 2v2 tactical shooter game, requiring only modest CPU hardware during inference. Our approach leverages an innovative, pixel-free perception architecture using a small set of ray-cast sensors, which capture essential spatial information efficiently. These sensors allow AI to perform competently without the computational overhead of traditional methods. Models are trained to mimic human behavior using supervised learning on human trajectory data, resulting in realistic and engaging AI agents. Human evaluation tests confirm that our AI agents provide human-like gameplay experiences while operating efficiently under computational constraints. This offers a significant advancement in AI model development for tactical shooter games and possibly other genres.

Solving complex real-world tasks requires cycles of actions and observations. This is particularly true in science, where tasks require many cycles of analysis, tool use, and experimentation. Language agents are promising for automating intellectual tasks in science because they can interact with tools via natural language or code. Yet their flexibility creates conceptual and practical challenges for software implementations, since agents may comprise non-standard components such as internal reasoning, planning, tool usage, as well as the inherent stochasticity of temperature-sampled language models. Here, we introduce Aviary, an extensible gymnasium for language agents. We formalize agents as policies solving language-grounded partially observable Markov decision processes, which we term language decision processes. We then implement five environments, including three challenging scientific environments: (1) manipulating DNA constructs for molecular cloning, (2) answering research questions by accessing scientific literature, and (3) engineering protein stability. These environments were selected for their focus on multi-step reasoning and their relevance to contemporary biology research. Finally, with online training and scaling inference-time compute, we show that language agents backed by open-source, non-frontier LLMs can match and exceed both frontier LLM agents and human experts on multiple tasks at up to 100x lower inference cost.

A deformation technique, known as the warped convolution, takes quantum fields in Minkowski spacetime to quantum fields in noncommutative Minkowski space-time. Since a quantum field is an operator valued regular distribution and the warped convolution is (weakly) an oscillatory integral of Rieffel type, we prove that the symbol classes introduced by Hormander admit extensions which are suited to the warped convolutions of scalar quantum field operators. We further show that, if a particular vector state on the undeformed algebra of field operators fulfills the microlocal spectrum condition, then every vector state on the deformed algebra generated by these warped convolutions fulfills the microlocal spectrum condition., Comment: 37 pages

We introduce a metric on the set of permutations of given order, which is a weighted generalization of Kendall's $\tau$ rank distance and study its properties. Using the edge graph of a permutohedron, we give a criterion which guarantees that a permutation lies metrically between another two fixed permutations. In addition, the conditions under which four points from the resulting metric space form a pseudolinear quadruple were found., Comment: 17 pages, 3 figures

General matrix multiplication (GEMM) is a ubiquitous computing kernel/algorithm for data processing in diverse applications, including artificial intelligence (AI) and deep learning (DL). Recent shift towards edge computing has inspired GEMM architectures based on unary computing, which are predominantly stochastic and rate-coded systems. This paper proposes a novel GEMM architecture based on temporal-coding, called tuGEMM, that performs exact computation. We introduce two variants of tuGEMM, serial and parallel, with distinct area/power-latency trade-offs. Post-synthesis Power-Performance-Area (PPA) in 45 nm CMOS are reported for 2-bit, 4-bit, and 8-bit computations. The designs illustrate significant advantages in area-power efficiency over state-of-the-art stochastic unary systems especially at low precisions, e.g. incurring just 0.03 mm^2 and 9 mW for 4 bits, and 0.01 mm^2 and 4 mW for 2 bits. This makes tuGEMM ideal for power constrained mobile and edge devices performing always-on real-time sensory processing., Comment: Published in 2023 IEEE International Symposium on Circuits and Systems (ISCAS), Monterey, CA, USA, 2023

General Matrix Multiplication (GEMM) is a ubiquitous compute kernel in deep learning (DL). To support energy-efficient edge-native processing, new GEMM hardware units have been proposed that operate on unary encoded bitstreams using much simpler hardware. Most unary approaches thus far focus on rate-based unary encoding of values and perform stochastic approximate computation. This work presents tubGEMM, a novel matrix-multiply unit design that employs hybrid temporal-unary and binary (tub) encoding and performs exact (not approximate) GEMM. It intrinsically exploits dynamic value sparsity to improve energy efficiency. Compared to the current best unary design uGEMM, tubGEMM significantly reduces area, power, and energy by 89\%, 87\%, and 50\%, respectively. A tubGEMM design performing 128x128 matrix multiply on 8-bit integers, in commercial TSMC N5 (5nm) process node, consumes just 0.22 mm^2 die area, 417.72 mW power, and 8.86 uJ energy, assuming no sparsity. Typical sparsity in DL workloads (MobileNetv2, ResNet-50) reduces energy by more than 3x, and lowering precision to 4 and 2 bits further reduces it by 24x and 104x respectively., Comment: Published in 2023 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)

We investigate state designs for continuous-variable quantum systems using the aid of lattice-like quantum states. These are code states of Gottesman-Kitaev-Preskill (GKP) codes. We show that for an n-mode system, the set of all GKP states forms a rigged continuous-variable state 2-design. We use these lattice state designs to construct a continuous variable shadow tomography protocol, derive sample complexity bounds for both global- and local GKP shadows under reasonable physical assumptions, and provide the physical gadgets needed to implement this protocol., Comment: 7+31 pages, 3 figures, comments welcome! v2 contains corrections of minor errors

In this article, we address the metrological problem of estimating collective stochastic properties of a many-body quantum system. Canonical examples include center-of-mass quadrature fluctuations in a system of bosonic modes and correlated dephasing in an ensemble of qubits (e.g., spins) or fermions. We develop a theoretical framework to determine the limits of correlated (weak) noise estimation with quantum sensor networks and unveil the requirements for entanglement advantage. Notably, an advantage emerges from the synergistic interplay between quantum correlations of the sensors and classical (spatial) correlations of the noises. We determine optimal entangled probe states and identify a sensing protocol, reminiscent of a many-body echo sequence, that achieves the fundamental limits of measurement sensitivity for a broad class of problems., Comment: 7+6 pages, 1 figure. Feedback welcome!

In the era of exoplanet studies with JWST, the transiting, hot gas giant WASP-52 b provides an excellent target for atmospheric characterization through transit spectroscopy. WASP-52 b orbits an active K-type dwarf recognized for its surface heterogeneities, such as star-spots and faculae, which offers challenges to atmospheric characterization via transmission spectroscopy. Previous transit observations have detected active regions on WASP-52 through crossing events in transit light-curves and via the spectral imprint of unocculted magnetic regions on transmission spectra. Here, we present the first JWST observations of WASP-52 b. Our JWST NIRISS/SOSS transit observation, obtained through the GTO 1201 Program, detects two clear spot-crossing events that deform the 0.6-2.8 $\mu$m transit light-curves of WASP-52 b. We find that these two occulted spots combined cover about 2.4 % of the stellar surface and have temperatures about 400-500 K colder than the stellar photosphere. Our NIRISS/SOSS transmission spectrum is best-fit by an atmosphere with H$_2$O (10.8 $\sigma$), He (7.3 $\sigma$, with evidence of an escaping tail at $\sim$ 2.9 $\sigma$), hints of K (2.5 $\sigma$), and unocculted star-spots and faculae (3.6 $\sigma$). The retrieved H$_2$O abundance ($\log$ H$_2$O $\approx -4 \pm 1$) is consistent with a subsolar or solar atmospheric metallicity for two independent data reductions. Our results underscore the importance of simultaneously modelling planetary atmospheres and unocculted stellar heterogeneities when interpreting transmission spectra of planets orbiting active stars and demonstrate the necessity of considering different stellar contamination models that account for both cold and hot active regions., Comment: Submitted to MNRAS; 17 pages, 11 figures and 3 tables

We present astrometric results from a Hubble Space Telescope (HST) campaign aimed at determining precise distances for cold Y-type brown dwarfs. Combining observations from a dedicated HST/WFC3 program with archival data, we derive astrometric solutions for 15 nearby Y dwarfs, by linking the high-precision relative astrometry from Hubble to the high-accuracy Gaia DR3 absolute reference system, using stars present in both to anchor the two frames of reference. We reach uncertainties on parallaxes below the 1-mas level for half of the sample, and down to 3 mas for two thirds of the targets, or relative precisions <1% in most cases and 2-5x improvements over previous measurements. For the remaining targets, we achieved slightly lower precisions on parallaxes (5-12 mas, 5-10%), correlated with the lower signal-to-noise of the faintest targets. The precision reached in our derived proper motions is around 0.1-0.4 mas/yr for most targets, and up to 1-2 mas/yr for less precise cases. Our estimated parallaxes and proper motions are generally in good agreement with literature values, and consistent to 1-2 sigma with recent Spitzer-derived parallaxes in most cases. These new astrometric solutions provide important validation of these objects' distances and sky motions, especially given the large disparities seen in previous estimates. Our results demonstrate the power of HST combined with Gaia to measure highly-precise absolute astrometry for faint brown dwarfs, and highlights the limitations reached for the reddest and coldest objects, for which JWST will certainly provide a favourable platform to improve these results., Comment: Accepted in MNRAS, 17 pages, 19 figures

Myocardial perfusion imaging using SPECT is widely utilized to diagnose coronary artery diseases, but image quality can be negatively affected in low-dose and few-view acquisition settings. Although various deep learning methods have been introduced to improve image quality from low-dose or few-view SPECT data, previous approaches often fail to generalize across different acquisition settings, limiting their applicability in reality. This work introduced DiffSPECT-3D, a diffusion framework for 3D cardiac SPECT imaging that effectively adapts to different acquisition settings without requiring further network re-training or fine-tuning. Using both image and projection data, a consistency strategy is proposed to ensure that diffusion sampling at each step aligns with the low-dose/few-view projection measurements, the image data, and the scanner geometry, thus enabling generalization to different low-dose/few-view settings. Incorporating anatomical spatial information from CT and total variation constraint, we proposed a 2.5D conditional strategy to allow the DiffSPECT-3D to observe 3D contextual information from the entire image volume, addressing the 3D memory issues in diffusion model. We extensively evaluated the proposed method on 1,325 clinical 99mTc tetrofosmin stress/rest studies from 795 patients. Each study was reconstructed into 5 different low-count and 5 different few-view levels for model evaluations, ranging from 1% to 50% and from 1 view to 9 view, respectively. Validated against cardiac catheterization results and diagnostic comments from nuclear cardiologists, the presented results show the potential to achieve low-dose and few-view SPECT imaging without compromising clinical performance. Additionally, DiffSPECT-3D could be directly applied to full-dose SPECT images to further improve image quality, especially in a low-dose stress-first cardiac SPECT imaging protocol., Comment: 13 pages, 6 figures, 2 tables. Paper under review. Oral presentation at IEEE MIC 2024

As data volumes expand rapidly, distributed machine learning has become essential for addressing the growing computational demands of modern AI systems. However, training models in distributed environments is challenging with participants hold skew, Non-Independent-Identically distributed (Non-IID) data. Low-Rank Adaptation (LoRA) offers a promising solution to this problem by personalizing low-rank updates rather than optimizing the entire model, LoRA-enabled distributed learning minimizes computational and maximize personalization for each participant. Enabling more robust and efficient training in distributed learning settings, especially in large-scale, heterogeneous systems. Despite the strengths of current state-of-the-art methods, they often require manual configuration of the initial rank, which is increasingly impractical as the number of participants grows. This manual tuning is not only time-consuming but also prone to suboptimal configurations. To address this limitation, we propose AutoRank, an adaptive rank-setting algorithm inspired by the bias-variance trade-off. AutoRank leverages the MCDA method TOPSIS to dynamically assign local ranks based on the complexity of each participant's data. By evaluating data distribution and complexity through our proposed data complexity metrics, AutoRank provides fine-grained adjustments to the rank of each participant's local LoRA model. This adaptive approach effectively mitigates the challenges of double-imbalanced, non-IID data. Experimental results demonstrate that AutoRank significantly reduces computational overhead, enhances model performance, and accelerates convergence in highly heterogeneous federated learning environments. Through its strong adaptability, AutoRank offers a scalable and flexible solution for distributed machine learning.

Combining multiple machine learning models has long been a technique for enhancing performance, particularly in distributed settings. Traditional approaches, such as model ensembles, work well, but are expensive in terms of memory and compute. Recently, methods based on averaging model parameters have achieved good results in some settings and have gained popularity. However, merging models initialized differently that do not share a part of their training trajectories can yield worse results than simply using the base models, even after aligning their neurons. In this paper, we introduce a novel approach, Non-uniform Parameter-wise Model Merging, or NP Merge, which merges models by learning the contribution of each parameter to the final model using gradient-based optimization. We empirically demonstrate the effectiveness of our method for merging models of various architectures in multiple settings, outperforming past methods. We also extend NP Merge to handle the merging of multiple models, showcasing its scalability and robustness., Comment: 9 pages, 1 figure, to be published in the Proceedings of the 9th IEEE Special Session on Machine Learning on Big Data (MLBD 2024)

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.

This paper explores image modeling from the frequency space and introduces DCTdiff, an end-to-end diffusion generative paradigm that efficiently models images in the discrete cosine transform (DCT) space. We investigate the design space of DCTdiff and reveal the key design factors. Experiments on different frameworks (UViT, DiT), generation tasks, and various diffusion samplers demonstrate that DCTdiff outperforms pixel-based diffusion models regarding generative quality and training efficiency. Remarkably, DCTdiff can seamlessly scale up to high-resolution generation without using the latent diffusion paradigm. Finally, we illustrate several intriguing properties of DCT image modeling. For example, we provide a theoretical proof of why `image diffusion can be seen as spectral autoregression', bridging the gap between diffusion and autoregressive models. The effectiveness of DCTdiff and the introduced properties suggest a promising direction for image modeling in the frequency space. The code is at \url{https://github.com/forever208/DCTdiff}., Comment: 23 pages

Exchange-only (EO) spin qubits in quantum dots offer an expansive design landscape for architecting scalable device layouts. The study of two-EO-qubit operations, which involve six electrons in six quantum dots, has so far been limited to a small number of the possible configurations, and previous works lack analyses of design considerations and implications for quantum error correction. Using a simple and fast optimization method, we generate complete pulse sequences for CX, CZ, iSWAP, leakage-controlled CX, and leakage-controlled CZ two-qubit gates on 450 unique planar six-dot topologies and analyze differences in sequence length (up to 43% reduction) across topology classes. In addition, we show that relaxing constraints on post-operation spin locations can yield further reductions in sequence length; conversely, constraining these locations in a particular way generates a CXSWAP operation with minimal additional cost over a standard CX. We integrate this pulse library into the Intel quantum stack and experimentally verify pulse sequences on a Tunnel Falls chip for different operations in a linear-connectivity device to confirm that they work as expected. Finally, we explore architectural implications of these results for quantum error correction. Our work guides hardware and software design choices for future implementations of scalable quantum dot architectures., Comment: 15 pages, 12 figures

Radio frequency interference (RFI) is a persistent contaminant in terrestrial radio astronomy. While new radio interferometers are becoming operational, novel sources of RFI are also emerging. In order to strengthen the mitigation of RFI in modern radio interferometers, we propose an on-line RFI mitigation scheme that can be run in the correlator of such interferometers. We combine statistics based on the energy as well as the polarization alignment of the correlated signal to develop an on-line RFI mitigation scheme that can be applied to a data stream produced by the correlator in real-time, especially targeted at low duty-cycle or transient RFI detection. In order to improve the computational efficiency, we explore the use of both single precision and half precision floating point operations in implementing the RFI mitigation algorithm. This ideally suits its deployment in accelerator computing devices such as graphics processing units (GPUs) as used by the LOFAR correlator. We provide results based on real data to demonstrate the efficacy of the proposed method.

Discrete-step walks describe the dynamics of particles in a lattice subject to hopping or splitting events at discrete times. Despite being of primordial interest to the physics of quantum walks, the topological properties arising from their discrete-step nature have been hardly explored. Here we report the observation of topological phases unique to discrete-step walks. We use light pulses in a double-fibre ring setup whose dynamics maps into a two-dimensional lattice subject to discrete splitting events. We show that the number of edge states is not simply described by the bulk invariants of the lattice (i.e., the Chern number and the Floquet winding number) as would be the case in static lattices and in lattices subject to smooth modulations. The number of edge states is also determined by a topological invariant associated to the discrete-step unitary operators acting at the edges of the lattice. This situation goes beyond the usual bulk-edge correspondence and allows manipulating the number of edge states without the need to go through a gap closing transition. Our work opens new perspectives for the engineering of topological modes for particles subject to quantum walks., Comment: 6 pages, 5 figures, see ancillary files section of the arxiv page of the article

Functional magnetic resonance imaging (fMRI) has become instrumental in researching brain function. One application of fMRI is investigating potential neural features that distinguish people with autism spectrum disorder (ASD) from healthy controls. The Autism Brain Imaging Data Exchange (ABIDE) facilitates this research through its extensive data-sharing initiative. While ABIDE offers data preprocessed with various atlases, independent component analysis (ICA) for dimensionality reduction remains underutilized. We address this gap by presenting ICA-based resting-state networks (RSNs) from preprocessed scans from ABIDE, now publicly available: https://github.com/SjirSchielen/groupICAonABIDE. These RSNs unveil neural activation clusters without atlas constraints, offering a perspective on ASD analyses that complements the predominantly atlas-based literature. This contribution provides a valuable resource for further research into ASD, potentially aiding in developing new analytical approaches.

Scaling the input image resolution is essential for enhancing the performance of Vision Language Models (VLMs), particularly in text-rich image understanding tasks. However, popular visual encoders such as ViTs become inefficient at high resolutions due to the large number of tokens and high encoding latency caused by stacked self-attention layers. At different operational resolutions, the vision encoder of a VLM can be optimized along two axes: reducing encoding latency and minimizing the number of visual tokens passed to the LLM, thereby lowering overall latency. Based on a comprehensive efficiency analysis of the interplay between image resolution, vision latency, token count, and LLM size, we introduce FastVLM, a model that achieves an optimized trade-off between latency, model size and accuracy. FastVLM incorporates FastViTHD, a novel hybrid vision encoder designed to output fewer tokens and significantly reduce encoding time for high-resolution images. Unlike previous methods, FastVLM achieves the optimal balance between visual token count and image resolution solely by scaling the input image, eliminating the need for additional token pruning and simplifying the model design. In the LLaVA-1.5 setup, FastVLM achieves 3.2$\times$ improvement in time-to-first-token (TTFT) while maintaining similar performance on VLM benchmarks compared to prior works. Compared to LLaVa-OneVision at the highest resolution (1152$\times$1152), FastVLM achieves comparable performance on key benchmarks like SeedBench and MMMU, using the same 0.5B LLM, but with 85$\times$ faster TTFT and a vision encoder that is 3.4$\times$ smaller.

Limitations in measurement instruments can hinder the implementation of some quantum algorithms. Understanding the suboptimality of such measurements with restrictions may then lead to more efficient measurement policies. In this paper, we theoretically examine the suboptimality arising from a Procrustes problem for minimizing the average distance between two fixed quantum states when one of the states has been measured by a Projective Measurement (PM). Specifically, we compare optima when we can only use PMs that are aligned with a low-rank subspace where the quantum states are supported, and when we can measure with the full set of PMs. For this problem, we show that the suboptimality ratio is independent of the dimension of the full space, and is at most polylogarithmic in the dimension of the low-rank subspace. In the proof of this result, we use a probabilistic approach and the main techniques include trace inequalities related to projective measurements, and operator norm bounds for equipartitions of Parseval frames, which are of independent interest., Comment: 36 pages, 3 figures

We present an application of a single-qubit Data Re-Uploading (QRU) quantum model for particle classification in calorimetric experiments. Optimized for Noisy Intermediate-Scale Quantum (NISQ) devices, this model requires minimal qubits while delivering strong classification performance. Evaluated on a novel simulated dataset specific to particle physics, the QRU model achieves high accuracy in classifying particle types. Through a systematic exploration of model hyperparameters -- such as circuit depth, rotation gates, input normalization and the number of trainable parameters per input -- and training parameters like batch size, optimizer, loss function and learning rate, we assess their individual impacts on model accuracy and efficiency. Additionally, we apply global optimization methods, uncovering hyperparameter correlations that further enhance performance. Our results indicate that the QRU model attains significant accuracy with efficient computational costs, underscoring its potential for practical quantum machine learning applications., Comment: 17 pages, 22 figures

While the existence of an asymmetry between the two Maxwell centroids at the centre of the two foveas recorded using a foveascope, leads to the ocular dominance in good readers, the lack of asymmetry in most of the observers with dyslexia leads to their non-dominance and their difficulties in reading and writing. Indeed, the lack of asymmetry between the two main roads to the brain, i.e. the two optical nerves, leads to perturbations in the brain central connectivity, namely between the two hemispheres inducing too robust interhemispheric visual connections, beyond the critical period of 7-8 years. The symmetrical mirror-connections like b-d (observed for about 60% of children with dyslexia) or the non-symmetrical connections like b-b (35%) induce confusions and duplications in observers with dyslexia but remain erasable thanks to pulsed systems (glasses, lamps, screens. . .), using Hebbian processes in primary cortex synapses. The effect is instantaneous, non-invasive and compensates the lack of asymmetry. These systems help most of the observers with dyslexia to overcome their difficulties. The orthoptists and speech therapists can acquire and improve the mechanism to help the children to catch up, and perhaps be able to use the foveascope for an early diagnosis with young children., Comment: 8 pages, 4 figures

Microdosimetry investigates the energy deposition of ionizing radiation at microscopic scales, beyond the assessment capabilities of macroscopic dosimetry. This contributes to an understanding of the biological response in radiobiology, radiation protection and radiotherapy. Microdosimetric pulse height spectra are usually measured using an ionization detector in a pulsed readout mode. This incorporates and a charge-sensitive amplifier followed by a shaping network. At high particle rates, the pileup of multiple pulses leads to distortions in the recorded spectra. Especially for gas-based detectors, this is a significant issue, that can be reduced by using solid-state detectors with smaller cross-sectional areas and faster readout speeds. At particle rates typical for ion therapy, however, such devices will also experience pileup. Mitigation techniques often focus on avoiding pileup altogether, while post-processing approaches are rarely investigated. This work explores pileup effects in microdosimetric measurements and presents a stochastic resampling algorithm, allowing for offline simulation and correction of spectra. Initially it was developed for measuring neutron spectra with tissue equivalent proportional counters and is adapted for the use with solid-state microdosimeters in a clinical radiotherapy setting. The algorithm was tested on data acquired with solid-state microdosimeters at the MedAustron ion therapy facility. The successful simulation and reduction of pileup counts is achieved by establishing of a limited number of parameters for a given setup. The presented results illustrate the potential of offline correction methods in situations where a direct pileup-free measurement is currently not practicable., Comment: Prepared for submission to Physics in Medicine & Biology Revised for improved readability

The main goal of present paper is to analyze the adiabatic definition of scattering matrix in the formalism of L-functionals. This definition leads to the notion of inclusive scattering matrix closely related to inclusive cross sections. We discuss this notion and the relation of our techniques to adiabatic quantum computing.

Many algorithms have been developed to estimate probability distributions subject to differential privacy (DP): such an algorithm takes as input independent samples from a distribution and estimates the density function in a way that is insensitive to any one sample. A recent line of work, initiated by Raskhodnikova et al. (Neurips '21), explores a weaker objective: a differentially private algorithm that approximates a single sample from the distribution. Raskhodnikova et al. studied the sample complexity of DP \emph{single-sampling} i.e., the minimum number of samples needed to perform this task. They showed that the sample complexity of DP single-sampling is less than the sample complexity of DP learning for certain distribution classes. We define two variants of \emph{multi-sampling}, where the goal is to privately approximate $m>1$ samples. This better models the realistic scenario where synthetic data is needed for exploratory data analysis. A baseline solution to \emph{multi-sampling} is to invoke a single-sampling algorithm $m$ times on independently drawn datasets of samples. When the data comes from a finite domain, we improve over the baseline by a factor of $m$ in the sample complexity. When the data comes from a Gaussian, Ghazi et al. (Neurips '23) show that \emph{single-sampling} can be performed under approximate differential privacy; we show it is possible to \emph{single- and multi-sample Gaussians with known covariance subject to pure DP}. Our solution uses a variant of the Laplace mechanism that is of independent interest. We also give sample complexity lower bounds, one for strong multi-sampling of finite distributions and another for weak multi-sampling of bounded-covariance Gaussians., Comment: 22 pages

Plastic scintillators are widely used for the detection of elementary particles, and 3D reconstruction of particle tracks is achieved by segmenting the detector into 3D granular structures. In this study, we present a novel prototype fabricated by additive manufacturing, consisting of a 5 x 5 x 5 array of 1 cm3 plastic scintillator cubes, each optically isolated. This innovative approach eliminates the need to construct complex monolithic geometries in a single operation and gets rid of the traditional time-consuming manufacturing and assembling processes. The prototype underwent performance characterization during a beam test at CERN's Proton-Synchrotron facility. Light yield, optical crosstalk, and light response uniformity, were evaluated. The prototype demonstrated a consistent light yield of approximately 27 photoelectrons (p.e.) per channel, similar to traditional cast scintillator detectors. Crosstalk between adjacent cubes averaged 4-5%, and light yield uniformity within individual cubes exhibited about 7% variation, indicating stability and reproducibility. These results underscore the potential of the novel additive manufacturing technique, for efficient and reliable production of high-granularity scintillator detectors.

The $p$-Laplacian evolution equation and the $\alpha$-Gauss curvature flow with a flat side are degenerate parabolic equations with evolving free boundaries. We give proofs of smooth short-time existence, up to the free boundaries, using a result of the authors on linear degenerate equations on a fixed domain., Comment: 15 pages. v2 conditions for Theorem 1.2 corrected

As learning-based methods for legged robots rapidly grow in popularity, it is important that we can provide safety assurances efficiently across different controllers and environments. Existing works either rely on a priori knowledge of the environment and safety constraints to ensure system safety or provide assurances for a specific locomotion policy. To address these limitations, we propose an observation-conditioned reachability-based (OCR) safety-filter framework. Our key idea is to use an OCR value network (OCR-VN) that predicts the optimal control-theoretic safety value function for new failure regions and dynamic uncertainty during deployment time. Specifically, the OCR-VN facilitates rapid safety adaptation through two key components: a LiDAR-based input that allows the dynamic construction of safe regions in light of new obstacles and a disturbance estimation module that accounts for dynamics uncertainty in the wild. The predicted safety value function is used to construct an adaptive safety filter that overrides the nominal quadruped controller when necessary to maintain safety. Through simulation studies and hardware experiments on a Unitree Go1 quadruped, we demonstrate that the proposed framework can automatically safeguard a wide range of hierarchical quadruped controllers, adapts to novel environments, and is robust to unmodeled dynamics without a priori access to the controllers or environments - hence, "One Filter to Deploy Them All". The experiment videos can be found on the project website., Comment: Project website: https://sia-lab-git.github.io/One_Filter_to_Deploy_Them_All/

We present an investigation into the effects of high-energy proton damage on charge trapping in germanium cross-strip detectors, with the goal of accomplishing three important measurements. First, we calibrated and characterized the spectral resolution of a spare COSI-balloon detector in order to determine the effects of intrinsic trapping, finding that electron trapping due to impurities dominates over hole trapping in the undamaged detector. Second, we performed two rounds of proton irradiation of the detector in order to quantify, for the first time, the rate at which charge traps are produced by proton irradiation. We find that the product of the hole trap density and cross-sectional area, $[n\sigma]_\mathrm{h}$ follows a linear relationship with the proton fluence, $F_\mathrm{p}$, with a slope of $(5.4\pm0.4)\times10^{-11}\,\mathrm{cm/p^{+}}$. Third, by utilizing our measurements of physical trapping parameters, we performed calibrations which corrected for the effects of trapping and mitigated degradation to the spectral resolution of the detector., Comment: 17 pages, 8 figures, 3 tables; submitted to Experimental Astronomy

Math reasoning is becoming an ever increasing area of focus as we scale large language models. However, even the previously-toughest evals like MATH are now close to saturated by frontier models (90.0% for o1-mini and 86.5% for Gemini 1.5 Pro). We introduce HARP, Human Annotated Reasoning Problems (for Math), consisting of 5,409 problems from the US national math competitions (A(J)HSME, AMC, AIME, USA(J)MO). Of these, 4,780 have answers that are automatically check-able (with libraries such as SymPy). These problems range six difficulty levels, with frontier models performing relatively poorly on the hardest bracket of 197 problems (average accuracy 41.1% for o1-mini, and 9.6% for Gemini 1.5 Pro). Our dataset also features multiple choices (for 4,110 problems) and an average of two human-written, ground-truth solutions per problem, offering new avenues of research that we explore briefly. We report evaluations for many frontier models and share some interesting analyses, such as demonstrating that frontier models across families intrinsically scale their inference-time compute for more difficult problems. Finally, we open source all code used for dataset construction (including scraping) and all code for evaluation (including answer checking) to enable future research at: https://github.com/aadityasingh/HARP., Comment: 28 pages, 17 figures

Magnetic fields remain one of the least understood aspects of exoplanetary systems. A deeper understanding of planetary dynamos and the evolution of surface magnetic properties throughout a planet's lifetime is a key scientific purpose, with implications for planetary evolution, habitability, and atmospheric dynamics. This study models the evolution of magnetic fields generated by dynamo action in cold giant gaseous planets. We solve the resistive magnetohydrodynamic (MHD) equations under anelastic approximation with a 3D pseudo-spectral spherical shell MHD code. We employ 1D thermodynamical hydrostatic profiles taken from gas giant evolutionary models as the background states of our MHD models. Numerical integration leads to saturated dynamo solutions. Such calculations are performed with radial profiles corresponding to different planetary ages so that we can interpret them as different snapshots of the magnetoconvection evolution during the long-term planetary evolution. We characterize magnetic fields across different evolutionary stages of a cold gaseous planet in terms of topology and strength. We find the occurrence of a transition from multipolar to dipolar-dominated dynamo regime throughout the life of a Jovian planet. During the planetary evolution and the cooling down phase, we observe a decrease in the average magnetic field strength near the dynamo surface as $\sim t^{-0.2}-t^{-0.3}$, a trend compatible with previously proposed scaling laws. We also find that some dimensionless parameters evolve differently for the multipolar to dipolar branch, possibly reflecting a force balance change. This approach can be extended to study hot gaseous planets, offering a versatile tool for interpreting the magnetic properties of giant planets., Comment: 19 pages, submitted for publication, comments are welcome

Dynamic quantum circuits incorporate mid-circuit measurements and feed-forward operations originally intended to realize Quantum Error Correction. This paradigm has recently been utilized to prepare certain states and long-range entangling gates as well as reduce resource overhead in quantum algorithms such as Quantum Fourier Transformation and Quantum Phase Estimation. In this paper, we present a novel framework for generating dynamic quantum circuits that automatically prepare any state or unitary operator. This procedure is powered by numerical optimization-based circuit synthesis methods. The first contribution is introducing optimization objective functions incorporating mid-circuit measurement and feed-forward operations. The second contribution is incorporating these into a popular open-source quantum circuit synthesis framework. We demonstrate the generation of dynamic circuits for state preparation, long-range entangling gates, circuit optimization, and the application of dynamic circuits to lattice simulations. The resulting circuits are validated through simulation and execution on quantum hardware. Furthermore, we perform noise analysis to explore the impact of different error ratios in mid-circuit measurements and gate errors, identifying scenarios where dynamic circuits offer the most significant benefits. The dynamic circuits generated by our framework show substantial improvements in reducing circuit depth and, in some cases, the number of gates required. To our knowledge, this is the first practical procedure to generate dynamic quantum circuits. Our objective functions are independent of the underlying synthesis framework and can be easily reused. This framework opens new possibilities for circuit generation and optimization methods, highlighting the potential of dynamic circuits to enhance the performance of quantum algorithms on near-term quantum computers.

This paper introduces Mayfly, a federated analytics approach enabling aggregate queries over ephemeral on-device data streams without central persistence of sensitive user data. Mayfly minimizes data via on-device windowing and contribution bounding through SQL-programmability, anonymizes user data via streaming differential privacy (DP), and mandates immediate in-memory cross-device aggregation on the server -- ensuring only privatized aggregates are revealed to data analysts. Deployed for a sustainability use case estimating transportation carbon emissions from private location data, Mayfly computed over 4 million statistics across more than 500 million devices with a per-device, per-week DP $\varepsilon = 2$ while meeting strict data utility requirements. To achieve this, we designed a new DP mechanism for Group-By-Sum workloads leveraging statistical properties of location data, with potential applicability to other domains., Comment: 22 pages, 7 figures

Pulse profile stability is a central assumption of standard pulsar timing methods. Thus, it is important for pulsar timing array experiments such as the North American Nanohertz Observatory for Gravitational Waves (NANOGrav) to account for any pulse profile variability present in their data sets. We show that in the NANOGrav 15-yr data set, the integrated pulse profile of PSR J1022+1001 as seen by the Arecibo radio telescope at 430, 1380, and 2030 MHz varies considerably in its shape from observation to observation. We investigate the possibility that this is due to the "ideal feed assumption" (IFA), on which NANOGrav's routine polarization calibration procedure relies. PSR J1022+1001 is $\sim 90\%$ polarized in one pulse profile component, and also has significant levels of circular polarization. Time-dependent deviations in the feed's polarimetric response (PR) could cause mixing between the intensity I and the other Stokes parameters, leading to the observed variability. We calibrate the PR using a mixture of Measurement Equation Modeling and Measurement Equation Template Matching techniques. The resulting profiles are no less variable than those calibrated using the IFA method, nor do they provide an improvement in the timing quality of this pulsar. We observe the pulse shape in 25-MHz bandwidths to vary consistently across the band, which cannot be explained by interstellar scintillation in combination with profile evolution with frequency. Instead, we favor phenomena intrinsic to the pulsar as the cause., Comment: 18 pages, 16 figures, 4 tables. Submitted to ApJ

Dipolar Bose-Einstein condensates are excellent platforms for studying supersolidity, characterized by coexisting density modulation and superfluidity. The realization of dipolar mixtures opens intriguing new scenarios, most remarkably the possibility of realizing a double supersolid, composed by two interacting superfluids. We analyze the complex excitation spectrum of a miscible trapped dipolar Bose mixture, showing that it provides key insights about the double supersolid regime. We show that this regime may be readily probed experimentally by monitoring the appearance of a doublet of superfluid compressional modes, linked to the different superfluid character of each component. Additionally, the dipolar supersolid mixture exhibits a non-trivial spin nature of the dipolar rotons, the Higgs excitation, and the low-lying Goldstone modes. Interestingly, the analysis of the lowest-lying modes allows for monitoring the transition of just one of the components into the incoherent droplet regime, whereas the other remains coherent, highlighting their disparate superfluid properties., Comment: 8 pages, 7 figures

In this paper, we introduce a geometric framework to analyze memorization in diffusion models using the eigenvalues of the Hessian of the log probability density. We propose that memorization arises from isolated points in the learned probability distribution, characterized by sharpness in the probability landscape, as indicated by large negative eigenvalues of the Hessian. Through experiments on various datasets, we demonstrate that these eigenvalues effectively detect and quantify memorization. Our approach provides a clear understanding of memorization in diffusion models and lays the groundwork for developing strategies to ensure secure and reliable generative models

We experimentally demonstrate a source of squeezed light featuring simultaneous ponderomotive squeezing from two mechanical modes of an optomechanical system. We use ultra-coherent vibrational modes ($Q$ factors on the order of $10^{8}$) of a soft-clamped membrane placed in a Fabry-P\'erot optical cavity at cryogenic conditions ($T=11\,\mathrm{K}$) and driven by quantum fluctuations in the intensity of light to create correlations between amplitude and phase quadratures of the intra-cavity light field. Continuous optical monitoring was conducted on two different mechanical modes with a frequency separation of around $1\,\mathrm{MHz}$. As a result of the interaction between the membrane position and the light, we generated ponderomotive squeezing of $4.8\,\mathrm{dB}$ for the first localized mechanical mode at $1.32\,\mathrm{MHz}$ and $4.2\,\mathrm{dB}$ for the second localized mode at $2.43\,\mathrm{MHz}$, as observed in direct detection when correcting for the detection inefficiency. Thus, we have demonstrated how squeezed light generation can be extended beyond a single octave in an optomechanical system by leveraging more than one mechanical mode. Utilizing homodyne detection to detect squeezing in an optimal quadrature would lead to squeezing levels at the output of the cavity of $7.3\,\mathrm{dB}$ and $6.8\,\mathrm{dB}$, in the two modes respectively. Squeezing of light demonstrated here for near-infrared light can be achieved in a broad range of wavelengths due to the relative insensitivity of optomechanical interaction with SiN membranes to the wavelength., Comment: 9 pages (+ 6 pages of Supplementary Material)

Critique between peers plays a vital role in the production of scientific knowledge. Yet, there is limited empirical evidence on the origins of criticism, its effects on the papers and individuals involved, and its visibility within the scientific literature. Here, we address these gaps through a data-driven analysis of papers that received substantiated and explicit written criticisms. Our analysis draws on data representing over 3,000 ``critical letters'' -- papers explicitly published to critique another -- from four high profile journals, with each letter linked to its target paper. We find that the papers receiving critical letters are disproportionately among the most highly-cited in their respective journal and, to a lesser extent, among the most interdisciplinary and novel. However, despite the theoretical importance of criticism in scientific progress, we observe no evidence that receiving a critical letter affects a paper's citation trajectory or the productivity and citation impact of its authors. One explanation for the limited consequence of critical letters is that they often go unnoticed. Indeed, we find that critical letters attract only a small fraction of the citations received by their targets, even years after publication. An analysis of topical similarity between criticized papers and their citing papers indicates that critical letters are primarily cited by researchers actively engaged in a similar field of study, whereas they are overlooked by more distant communities. Although criticism is celebrated as a cornerstone to science, our findings reveal that it is concentrated on high-impact papers, has minimal measurable consequences, and suffers from limited visibility. These results raise important questions about the role and value of critique in scientific practice., Comment: 27 pages, 4 figures

We discuss the H\"older regularity of solutions to the semilinear equation involving the fractional Laplacian $(-\Delta)^s u=f(u)$ in one dimension. We put in evidence a new regularity phenomenon which is a combined effect of the nonlocality and the semilinearity of the equation, since it does not happen neither for local semilinear equations, nor for nonlocal linear equations. Namely, for nonlinearities $f$ in $C^\beta$ and when $2s+\beta <1$, the solution is not always $C^{2s+\beta-\epsilon}$ for all $\epsilon >0$. Instead, in general the solution $u$ is at most $C^{2s/(1-\beta)}.$, Comment: 18 pages

This paper introduces a novel deconvolution algorithm, shift-invariant multi-linearity (SIML), which significantly enhances the analysis of data from a comprehensive two-dimensional gas chromatograph coupled to a mass spectrometric detector (GC$\times$GC-TOFMS). Designed to address the challenges posed by retention time shifts and high noise levels, SIML incorporates wavelet-based smoothing and Fourier-Transform based shift-correction within the multivariate curve resolution-alternating least squares (MCR-ALS) framework. We benchmarked the SIML algorithm against traditional methods such as MCR-ALS and Parallel Factor Analysis 2 with flexible coupling (PARAFAC2$\times$N) using both simulated and real GC$\times$GC-TOFMS datasets. Our results demonstrate that SIML provides unique solutions with significantly improved robustness, particularly in low signal-to-noise ratio scenarios, where it maintains high accuracy in estimating mass spectra and concentrations. The enhanced reliability of quantitative analyses afforded by SIML underscores its potential for broad application in complex matrix analyses across environmental science, food chemistry, and biological research., Comment: 27 pages pages

Auditing Differentially Private Stochastic Gradient Descent (DP-SGD) in the final model setting is challenging and often results in empirical lower bounds that are significantly looser than theoretical privacy guarantees. We introduce a novel auditing method that achieves tighter empirical lower bounds without additional assumptions by crafting worst-case adversarial samples through loss-based input-space auditing. Our approach surpasses traditional canary-based heuristics and is effective in both white-box and black-box scenarios. Specifically, with a theoretical privacy budget of $\varepsilon = 10.0$, our method achieves empirical lower bounds of $6.68$ in white-box settings and $4.51$ in black-box settings, compared to the baseline of $4.11$ for MNIST. Moreover, we demonstrate that significant privacy auditing results can be achieved using in-distribution (ID) samples as canaries, obtaining an empirical lower bound of $4.33$ where traditional methods produce near-zero leakage detection. Our work offers a practical framework for reliable and accurate privacy auditing in differentially private machine learning., Comment: 13 pages, 8 figures, NeurIPS (SFLLM Workshop)

Air pollution is a pressing environmental risk to public health, particularly in cities where population density and pollution levels are high. Traditional methods for exposure analysis often rely on census data, but recent studies highlight the impact of daily mobility on individuals' exposure. Here, we develop a methodology to determine unprecedented pedestrian exposure estimates at the city scale by combining sidewalk pedestrian flows with high-resolution (25 m x 25 m) NO2 data from bias-corrected predictions of the air quality system CALIOPE-Urban. Applied to Barcelona (Spain) for the year 2019, we show that pedestrian flow and NO2 levels exhibit negligible temporal correlation. While short-term (hourly) exposure is driven by pedestrian mobility, long-term (monthly) exposure is dominated by NO2 patterns. We identify strong spatial gradients of exposure, highlighting the importance for high-resolution solutions at the sidewalks scale. Finally, we determine that exposure mitigation strategies should consider different citizen subgroups based on their mobility and preferred routes, as significant differences were found between residential and pedestrian exposure. Our results provide exposure indicators designed for city planners and policymakers, helping to prioritize mitigation measures where and when they are most needed.

In this work, we present a novel method for extensive multi-scale generative terrain modeling. At the core of our model is a cascade of superresolution diffusion models that can be combined to produce consistent images across multiple resolutions. Pairing this concept with a tiled generation method yields a scalable system that can generate thousands of square kilometers of realistic Earth surfaces at high resolution. We evaluate our method on a dataset collected from Bing Maps and show that it outperforms super-resolution baselines on the extreme super-resolution task of 1024x zoom. We also demonstrate its ability to create diverse and coherent scenes via an interactive gigapixel-scale generated map. Finally, we demonstrate how our system can be extended to enable novel content creation applications including controllable world generation and 3D scene generation.