Search

This paper explores how generative AI can help automate and improve key steps in systems engineering. It examines AI's ability to analyze system requirements based on INCOSE's "good requirement" criteria, identifying well-formed and poorly written requirements. The AI does not just classify requirements but also explains why some do not meet the standards. By comparing AI assessments with those of experienced engineers, the study evaluates the accuracy and reliability of AI in identifying quality issues. Additionally, it explores AI's ability to classify functional and non-functional requirements and generate test specifications based on these classifications. Through both quantitative and qualitative analysis, the research aims to assess AI's potential to streamline engineering processes and improve learning outcomes. It also highlights the challenges and limitations of AI, ensuring its safe and ethical use in professional and academic settings., Comment: 4 pages, 2 figures

Two samples of silicon pad sensors and two samples of GaAs sensors are studied in an electron beam with 5 GeV energy from the DESY-II test-beam facility. The sizes of the silicon and GaAs sensors are about 9$\times$9 cm$^2$ and 5$\times$8 cm$^2$, respectively. The thickness is 500 micrometer for both the silicon and GaAs sensors. The pad size is about 5$\times$5 mm$^2$. The sensors are foreseen to be used in a compact electromagnetic sampling calorimeter. The readout of the pads is done by metal traces connected to the pads and the front-end ASICs at the edges of the sensors. For the silicon sensors, copper traces on a Kapton foil are connected to the sensor pads with conducting glue. The pads of the GaAs sensors are connected to bond-pads via aluminium traces on the sensor substrate. The readout is based on a dedicated front-end ASIC, called FLAME. Pre-processing of the raw data and deconvolution is performed with FPGAs. The whole system is orchestrated by a Trigger Logic Unit. Results are shown for the signal-to-noise ratio, the homogeneity of the response, edge effects on pads, and for signals due to the readout traces., Comment: 22 pages, 24 figures, submitted to The European Physical Journal C

Electrocardiogram (ECG)-based biometrics offer a promising method for user identification, combining intrinsic liveness detection with morphological uniqueness. However, elevated heart rates introduce significant physiological variability, posing challenges to pattern recognition systems and leading to a notable performance gap between resting and post-exercise conditions. Addressing this gap is critical for advancing ECG-based biometric systems for real-world applications. We propose DE-PADA, a Dual Expert model with Personalized Augmentation and Domain Adaptation, designed to enhance robustness across diverse physiological states. The model is trained primarily on resting-state data from the evaluation dataset, without direct exposure to their exercise data. To address variability, DE-PADA incorporates ECG-specific innovations, including heartbeat segmentation into the PQRS interval, known for its relative temporal consistency, and the heart rate-sensitive ST interval, enabling targeted feature extraction tailored to each region's unique characteristics. Personalized augmentation simulates subject-specific T-wave variability across heart rates using individual T-wave peak predictions to adapt augmentation ranges. Domain adaptation further improves generalization by leveraging auxiliary data from supplementary subjects used exclusively for training, including both resting and exercise conditions. Experiments on the University of Toronto ECG Database demonstrate the model's effectiveness. DE-PADA achieves relative improvements in post-exercise identification rates of 26.75% in the initial recovery phase and 11.72% in the late recovery phase, while maintaining a 98.12% identification rate in the sitting position. These results highlight DE-PADA's ability to address intra-subject variability and enhance the robustness of ECG-based biometric systems across diverse physiological states.

Most superconducting mechanisms pair electrons within the same band, forming spin singlets. However, the discovery of multi-band superconductivity has opened new scenarios for pairing, particularly in systems with strong spin-orbit coupling. Here, we reveal inter-band pairing in the superconductor \bipd\ by mapping the amplitude of sub-gap Yu-Shiba-Rusinov (YSR) states around Vanadium adatoms deposited on its surface. The surface of \bipd\ is characterized by spin-helical-like bands near the Fermi level. Scanning tunneling spectroscopy reveals anisotropic YSR amplitude oscillations around the impurity, driven by spin-conserving Bogoliubov quasiparticle interference (BQPI). Analysis of the BQPI patterns at the YSR energy exposes inter-band pairing in this material. Interestingly, only a small subset of all possible inter-band scattering processes observed in the normal state contribute to the BQPI patterns. Combining experimental data and theory, we demonstrate that the observed band selectivity results from the hybridization of the band coupled with the impurity with other bands. Our findings reveal unconventional pairing mechanisms in \bipd\ and highlight the crucial role of spin-orbit interactions in their formation., Comment: 6 pages, 5 figures

Generating novel crystalline materials has potential to lead to advancements in fields such as electronics, energy storage, and catalysis. The defining characteristic of crystals is their symmetry, which plays a central role in determining their physical properties. However, existing crystal generation methods either fail to generate materials that display the symmetries of real-world crystals, or simply replicate the symmetry information from examples in a database. To address this limitation, we propose SymmCD, a novel diffusion-based generative model that explicitly incorporates crystallographic symmetry into the generative process. We decompose crystals into two components and learn their joint distribution through diffusion: 1) the asymmetric unit, the smallest subset of the crystal which can generate the whole crystal through symmetry transformations, and; 2) the symmetry transformations needed to be applied to each atom in the asymmetric unit. We also use a novel and interpretable representation for these transformations, enabling generalization across different crystallographic symmetry groups. We showcase the competitive performance of SymmCD on a subset of the Materials Project, obtaining diverse and valid crystals with realistic symmetries and predicted properties.

Metalenses have become ubiquitous in academic research and have begun to make their transition to industry. However, chromatic aberration still greatly limits the applications of metalenses. Achieving a wide field-of-view (FOV) is another challenge that has been dealt with successfully by using a removed stop, but when combined with broadband spectrum, lateral chromatic aberration severely limits performance. In this paper we tackle this grand challenge and present a comprehensive design methodology for a simultaneously wide-FOV and achromatic metalens which is inspired by the human visual system. As a design example, we present a metalens operating in the near infrared (NIR), with 10% relative spectral bandwidth (807-893nm), focal length of 5mm, F/5, and FOV of +/-20 degrees. In particular, we show how to optimize the stop position, and correct the lateral chromatic aberration, both of which have not been reported in the past. In addition, we evaluate the performance of the metalens using accurate performance metrics, and demonstrate the improvement compared to a chromatic metalens. Our approach paves the way for the design of wide FOV metalenses that can operate over a relatively large bandwidth, effectively contributing to the widespread implementation of metalens science and technology.

The emergence of foundational models has greatly improved performance across various downstream tasks, with fine-tuning often yielding even better results. However, existing fine-tuning approaches typically require access to model weights and layers, leading to challenges such as managing multiple model copies or inference pipelines, inefficiencies in edge device optimization, and concerns over proprietary rights, privacy, and exposure to unsafe model variants. In this paper, we address these challenges by exploring "Gray-box" fine-tuning approaches, where the model's architecture and weights remain hidden, allowing only gradient propagation. We introduce a novel yet simple and effective framework that adapts to new tasks using two lightweight learnable modules at the model's input and output. Additionally, we present a less restrictive variant that offers more entry points into the model, balancing performance with model exposure. We evaluate our approaches across several backbones on benchmarks such as text-image alignment, text-video alignment, and sketch-image alignment. Results show that our Gray-box approaches are competitive with full-access fine-tuning methods, despite having limited access to the model.

Breaking crystal symmetry is essential for engineering emissive double perovskite metal halides. The goal is to overcome their inherently indirect and disallowed optical transitions. Here we introduce a synthesis for silver - Cs2AgInCl6 two-dimensional hybrid nanoplate products that break the symmetry in two ways, their shape and their heterointerfaces. A comparative study between Cs2AgInCl6 nanocubes and nanoplates is presented to emphasize the difference in optical properties. A modified colloidal synthesis for Cs2AgInCl6 yields high-quality nanoplates with small lateral dimensions very different from the symmetric cubes. Each nanoplate is decorated with metallic silver nanoparticles, with diameters on the scale of the thickness of the perovskite nanoplate, forming significant heterointerfaces that further break symmetry. The Cs2AgInCl6 two-dimensional nanoplates also demonstrate facile transformation into larger crystalline nanosheets once deposited on substrates. We thus highlight those nanoplates as potential building blocks for assemblies of functional surfaces., Comment: The authors Sasha Khalfin and Noam Veber contributed equally

We report an unconventional resistive switching effect on high-density self-assembled Ag-nanowire networks tailored by a fuse-like operation. We propose a mechanism to rationalize the observed phenomenology by analyzing the electrical signatures before and after such a fusing. The explanation allows reconciling the results obtained in similar systems early adopted as transparent electrodes and the more recent attempts to use this type of substrate for in-materia computational operations. In addition to the usual analog nature of the available resistance states and the ability to tune internal weights, we show that these networks' sparsity and non-linear behavior are also attributes. Thus, the formerly exhibited nanowires' abilities to code synaptic behavior are complemented by neuronal features upon properly tuning the network density and the applied electrical protocol.

We present a study of new 7.7-11.3 $\mu$m data obtained with the James Webb Space Telescope Mid-InfraRed Instrument in the starburst galaxy M 82. In particular, we focus on the dependency of the integrated CO(1-0) line intensity on the MIRI-F770W and MIRI-F1130W filter intensities to investigate the correlation between CO content and the 7.7 and 11.3 $\mu$m features from polycyclic aromatic hydrocarbons (PAH) in M 82's outflows. To perform our analysis, we identify CO clouds using archival $^{12}$CO($J$=1-0) NOEMA moment 0 map within 2 kpc from the center of M 82, with sizes ranging between $\sim$21 and 270 pc; then, we compute the CO-to-PAH relations for the 306 validated CO clouds. On average, the power-law slopes for the two relations in M 82 are lower than what is seen in local main-sequence spirals. In addition, there is a moderate correlation between $I_{\rm CO(1-0)}$-$I_{\rm 7.7\mu m} /I_{\rm 11.3\mu m}$ for most of the CO cloud groups analyzed in this work. Our results suggest that the extreme conditions in M 82 translate into CO not tracing the full budget of molecular gas in smaller clouds, perhaps as a consequence of photoionization and/or emission suppression of CO molecules due to hard radiation fields from the central starburst.

Despite significant advancements in video large multimodal models (video-LMMs), achieving effective temporal grounding in long-form videos remains a challenge for existing models. To address this limitation, we propose Temporal Preference Optimization (TPO), a novel post-training framework designed to enhance the temporal grounding capabilities of video-LMMs through preference learning. TPO adopts a self-training approach that enables models to differentiate between well-grounded and less accurate temporal responses by leveraging curated preference datasets at two granularities: localized temporal grounding, which focuses on specific video segments, and comprehensive temporal grounding, which captures extended temporal dependencies across entire video sequences. By optimizing on these preference datasets, TPO significantly enhances temporal understanding while reducing reliance on manually annotated data. Extensive experiments on three long-form video understanding benchmarks--LongVideoBench, MLVU, and Video-MME--demonstrate the effectiveness of TPO across two state-of-the-art video-LMMs. Notably, LLaVA-Video-TPO establishes itself as the leading 7B model on the Video-MME benchmark, underscoring the potential of TPO as a scalable and efficient solution for advancing temporal reasoning in long-form video understanding. Project page: https://ruili33.github.io/tpo_website.

Understanding the transition from atomic gas to molecular gas is critical to explain the formation and evolution of molecular clouds. However, the gas phases involved, cold HI and CO-dark molecular gas, are challenging to directly observe and physically characterize. We observed the Cygnus X star-forming complex in carbon radio recombination lines (CRRLs) at 274--399 MHz with the Green Bank Telescope at 21 pc (48') resolution. Of the 30 deg^2 surveyed, we detect line-synthesized C273alpha emission from 24 deg^2 and produce the first large-area maps of low-frequency CRRLs. The morphology of the C273alpha emission reveals arcs, ridges, and extended possibly sheet-like gas which are often on the outskirts of CO emission and likely transitioning from HI-to-H_2. The typical angular separation of C273alpha and 13CO emission is 12 pc, and we estimate C273alpha gas densities of n_H ~ 40 - 400 cm^3. The C273alpha line profiles are Gaussian and likely turbulent broadened, spanning a large range of FWHM from 2 to 20 km/s with a median of 10.6 km/s. Mach numbers fall within 10--30. The turbulent timescale is relatively short, 2.6 Myr, and we deduce that the turbulent pressure likely dominates the evolution of the C273alpha gas. Velocity offsets between C273alpha and 13CO components are apparent throughout the region and have a typical value of 2.9 km/s. Two regimes have emerged from the data: one regime in which C273alpha and 13CO are strongly related (at N_H ~ 4 x 10^21 cm^-2), and a second, in which C273alpha emits independently of the 13CO intensity. In the former regime, C273alpha may arise from the the envelopes of massive clouds (filaments), and in the latter, C273alpha emits from cold clumps in a more-diffuse mix of HI and H_2 gas., Comment: 26 pages. 13 figures. submitted to ApJ. Comments welcome!

Large language models (LLMs) have excelled in various applications, yet serving them at scale is challenging due to their substantial resource demands and high latency. Our real-world studies reveal that over 60% of user requests to LLMs have semantically similar counterparts, suggesting the potential for knowledge sharing among requests. However, naively caching and reusing past responses leads to large quality degradation. In this paper, we introduce EchoLM, an in-context caching system that leverages historical requests as examples to guide response generation, enabling selective offloading of requests to more efficient LLMs. However, enabling this real-time knowledge transfer leads to intricate tradeoffs between response quality, latency, and system throughput at scale. For a new request, EchoLM identifies similar, high-utility examples and efficiently prepends them to the input for better response. At scale, EchoLM adaptively routes requests to LLMs of varying capabilities, accounting for response quality and serving loads. EchoLM employs a cost-aware cache replay mechanism to improve example quality and coverage offline, maximizing cache utility and runtime efficiency. Evaluations on millions of open-source requests demonstrate that EchoLM has a throughput improvement of 1.4-5.9x while reducing latency by 28-71% without hurting response quality on average.

We present a method based on optimal transport to remove arbitrage opportunities within a finite set of option prices. The method is notably intended for regulatory stress-tests, which impose to apply important local distortions to implied volatility surfaces. The resulting stressed option prices are naturally associated to a family of signed marginal measures: we formulate the process of removing arbitrage as a projection onto the subset of martingale measures with respect to a Wasserstein metric in the space of signed measures. We show how this projection problem can be recast as an optimal transport problem; in view of the numerical solution, we apply an entropic regularization technique. For the regularized problem, we derive a strong duality formula, show convergence results as the regularization parameter approaches zero, and formulate a multi-constrained Sinkhorn algorithm, where each iteration involves, at worse, finding the root of an explicit scalar function. The convergence of this algorithm is also established. We compare our method with the existing approach by [Cohen, Reisinger and Wang, Appl.\ Math.\ Fin.\ 2020] across various scenarios and test cases.

In this work, we present first $^{236}$U results in the northwestern Mediterranean. $^{236}$U is studied in a seawater column sampled at DYFAMED (Dynamics of Atmospheric Fluxes in the Mediterranean Sea) station (Ligurian Sea). The obtained $^{236}$U/$^{238}$U atom ratios in the dissolved phase, ranging from about $2\times10^{-9}$ at $100$ m depth to about $1.5\times10^{-9}$ at $2350$ m depth, indicate that anthropogenic $^{236}$U dominates the whole water column. The corresponding deep-water column inventory ($12.6$ ng/m$^2$ or $32.1\times10^{12}$ atoms/m$^2$) exceeds by a factor of $2.5$ the expected one for global fallout at similar latitudes ($5$ ng/m$^2$ or $13\times10^{12}$ atoms/m$^2$), evidencing the influence of local or regional $^{236}$U sources in the western Mediterranean basin. On the other hand, the input of $^{236}$U associated to Saharan dust outbreaks is evaluated. It is estimated an additional $^{236}$U annual deposition of about $0.2$ pg/m$^2$ based on the study of atmospheric particles collected in Monaco during different Saharan dust intrusions. The obtained results in the corresponding suspended solids collected at DYFAMED station indicate that about $64$% of that $^{236}$U stays in solution in seawater. Overall, this source accounts for about $0.1$% of the $^{236}$U inventory excess observed at DYFAMED station. The influence of the so-called Chernobyl fallout and the radioactive effluents produced by the different nuclear installations allocated to the Mediterranean basin, might explain the inventory gap, however, further studies are necessary to come to a conclusion about its origin.

A simple and fast radiochemical procedure for the sequential extraction of U, Np and Pu from small-volume seawater samples ($<10$ L) is presented. The method has been developed and optimized for the final determination of $^{236}$U, $^{237}$Np, $^{239}$Pu and $^{240}$Pu by Accelerator Mass Spectrometry (AMS). It is based on the use of $^{242}$Pu as tracer for both Np and Pu isotopes. Samples are pre-concentrated by Fe(OH)$_2$ co-precipitation. TEVA and UTEVA resins are used in a very simplified way for the final purification of the Pu+Np and U fractions, respectively. The radiochemical yields of the three elements have been investigated in detail by alpha spectrometry (AS) and gamma spectrometry (GS). The obtained results indicate high and robust chemical yields for the three elements and similar ones for Pu and Np. Furthermore, the use of $^{242}$Pu as tracer for $^{237}$Np is validated by analyzing a reference seawater sample (IAEA-443) by radiometric techniques. We demonstrated that, if chemicals are properly chosen, processed blank levels can be kept at the same level of the extremely low detection limits that can be achieved by AMS ($10^5 - 10^6$ atoms/sample). The procedure is finally applied for the study by AMS of a reference seawater sample from the Mediterranean Sea (IAEA-418).

We address the challenges of Byzantine-robust training in asynchronous distributed machine learning systems, aiming to enhance efficiency amid massive parallelization and heterogeneous computing resources. Asynchronous systems, marked by independently operating workers and intermittent updates, uniquely struggle with maintaining integrity against Byzantine failures, which encompass malicious or erroneous actions that disrupt learning. The inherent delays in such settings not only introduce additional bias to the system but also obscure the disruptions caused by Byzantine faults. To tackle these issues, we adapt the Byzantine framework to asynchronous dynamics by introducing a novel weighted robust aggregation framework. This allows for the extension of robust aggregators and a recent meta-aggregator to their weighted versions, mitigating the effects of delayed updates. By further incorporating a recent variance-reduction technique, we achieve an optimal convergence rate for the first time in an asynchronous Byzantine environment. Our methodology is rigorously validated through empirical and theoretical analysis, demonstrating its effectiveness in enhancing fault tolerance and optimizing performance in asynchronous ML systems.

Our present contribution sets out to investigate a scenario based on the effects of the Loop Quantum Gravity (LQG) on the electromagnetic sector of the Standard Model of Fundamental Interactions and Particle Physics (SM). Starting then from a post-Maxwellian version of Electromagnetism that includes LQG effects, we work out and discuss the influence of LQG parameters on classical quantities, such as the components of the stress-tensor. Furthermore, we inspect the propagation of electromagnetic waves and study optical properties of the QED vacuum in this scenario. Among these, we contemplate the combined effect between the LQG parameters and a homogeneous background magnetic field on the propagation of electromagnetic waves, considering in detail issues like group velocities and refractive indices of the QED vacuum. Finally, with the help of the LQG-extended photonic dispersion relations previously analyzed, we re-discuss the kinematics of the Compton effect and conclude that there emerges an interesting nonlinear profile in the wavelengths of both the incoming and the deflected photons., Comment: 9 pages, 0 figures

Generative modeling of symmetric densities has a range of applications in AI for science, from drug discovery to physics simulations. The existing generative modeling paradigm for invariant densities combines an invariant prior with an equivariant generative process. However, we observe that this technique is not necessary and has several drawbacks resulting from the limitations of equivariant networks. Instead, we propose to model a learned slice of the density so that only one representative element per orbit is learned. To accomplish this, we learn a group-equivariant canonicalization network that maps training samples to a canonical pose and train a non-equivariant generative model over these canonicalized samples. We implement this idea in the context of diffusion models. Our preliminary experimental results on molecular modeling are promising, demonstrating improved sample quality and faster inference time., Comment: NeurReps 2024 Workshop Version

The development of vision-language models (VLMs) is driven by large-scale and diverse multimodal datasets. However, progress toward generalist biomedical VLMs is limited by the lack of annotated, publicly accessible datasets across biology and medicine. Existing efforts are restricted to narrow domains, missing the full diversity of biomedical knowledge encoded in scientific literature. To address this gap, we introduce BIOMEDICA, a scalable, open-source framework to extract, annotate, and serialize the entirety of the PubMed Central Open Access subset into an easy-to-use, publicly accessible dataset. Our framework produces a comprehensive archive with over 24 million unique image-text pairs from over 6 million articles. Metadata and expert-guided annotations are also provided. We demonstrate the utility and accessibility of our resource by releasing BMCA-CLIP, a suite of CLIP-style models continuously pre-trained on the BIOMEDICA dataset via streaming, eliminating the need to download 27 TB of data locally. On average, our models achieve state-of-the-art performance across 40 tasks - spanning pathology, radiology, ophthalmology, dermatology, surgery, molecular biology, parasitology, and cell biology - excelling in zero-shot classification with a 6.56% average improvement (as high as 29.8% and 17.5% in dermatology and ophthalmology, respectively), and stronger image-text retrieval, all while using 10x less compute. To foster reproducibility and collaboration, we release our codebase and dataset for the broader research community.

The rapid development of vision language models (VLMs) demands rigorous and reliable evaluation. However, current visual question answering (VQA) benchmarks often depend on open-ended questions, making accurate evaluation difficult due to the variability in natural language responses. To address this, we introduce AutoConverter, an agentic framework that automatically converts these open-ended questions into multiple-choice format, enabling objective evaluation while reducing the costly question creation process. Our experiments demonstrate that AutoConverter can generate correct and challenging multiple-choice questions, with VLMs demonstrating consistently similar or lower accuracy on these questions compared to human-created ones. Using AutoConverter, we construct VMCBench, a benchmark created by transforming 20 existing VQA datasets into a unified multiple-choice format, totaling 9,018 questions. We comprehensively evaluate 33 state-of-the-art VLMs on VMCBench, setting a new standard for scalable, consistent, and reproducible VLM evaluation., Comment: Project page: https://yuhui-zh15.github.io/AutoConverter-Website/

The Kicked Rotor is perhaps the simplest physical model to illuminate the transition from regular to chaotic motion in classical mechanics. It is also widely applied as a model of light-matter interactions. In the conventional treatment, the infinitesimal width of each kick allows an immediate integration of the equations of motion. This in turn allows a full description of the dynamics via a discrete mapping, the Standard Map, if one looks at the dynamics only stroboscopically. It turns out that this model is only part of a much richer story if one accounts for finite temporal width of the kick. In this letter, we formulate a general model of finite-width periodic forcing and derive a continuous set of maps that depend on a parameter shift $\Delta$ that allows one to capture the motion in both the driven and kicked regimes. The fixed points and symmetry of the mapping are shown analytically and numerically to depend on the value of the shift parameter.

Autonomous high-speed navigation through large, complex environments requires real-time generation of agile trajectories that are dynamically feasible, collision-free, and satisfy state or actuator constraints. Most modern trajectory planning techniques rely on numerical optimization because high-quality, expressive trajectories that satisfy various constraints can be systematically computed. However, meeting computation time constraints and the potential for numerical instabilities can limit the use of optimization-based planners in safety-critical scenarios. This work presents an optimization-free planning framework that stitches short trajectory segments together with graph search to compute long range, expressive, and near-optimal trajectories in real-time. Our STITCHER algorithm is shown to outperform modern optimization-based planners through our innovative planning architecture and several algorithmic developments that make real-time planning possible. Extensive simulation testing is conducted to analyze the algorithmic components that make up STITCHER, and a thorough comparison with two state-of-the-art optimization planners is performed. It is shown STITCHER can generate trajectories through complex environments over long distances (tens of meters) with low computation times (milliseconds)., Comment: V1 Draft

Scintillation, the process of converting high-energy radiation to detectable visible light, is pivotal in advanced technologies spanning from medical diagnostics to fundamental scientific research. Despite significant advancements toward faster and more efficient scintillators, there remains a fundamental limit arising from the intrinsic properties of scintillating materials. The scintillation process culminates in spontaneous emission of visible light, which is restricted in rate by the oscillator strength of individual emission centers. Here, we observe a novel collective emission phenomenon under X-ray excitation, breaking this limit and accelerating the emission. Our observation reveals that strong interactions between simultaneously excited coupled perovskite quantum dots can create collective radioluminescence. This effect is characterized by a spectral shift and an enhanced rate of emission, with an average lifetime of 230 ps, 14 times faster than their room temperature spontaneous emission. It has been established that such quantum dots exhibit superfluorescence under UV excitation. However, X-ray superfluorescence is inherently different, as each high-energy photon creates multiple synchronized excitation events, triggered by a photoelectron and resulting in even faster emission rates, a larger spectral shift, and a broader spectrum. This observation is consistent with a quantum-optical analysis explaining both the UV-driven and X-ray-driven effects. We use a Hanbury-Brown-Twiss g^(2) ({\tau}) setup to analyze the temperature-dependent temporal response of these scintillators. Collective radioluminescence breaks the limit of scintillation lifetime based on spontaneous emission and could dramatically improve time-of-flight detector performance, introducing quantum enhancements to scintillation science.

Hanbury Brown and Twiss (HBT) interferometry is a milestone experiment that transformed our understanding of the nature of light. The concept was demonstrated in 1956 to measure the radii of stars through photon coincidence detection. This form of coincidence detection later became a cornerstone of modern quantum optics. Here we connect HBT interferometry to the physics of scintillation, the process of spontaneous light emission upon excitation by high-energy particles, such as x-rays. Our work reveals intrinsic photon bunching in the scintillation process, which we utilize to elucidate its underlying light emission mechanisms. Specifically, g^((2) ) ({\tau}) enables the quantitative extraction of scintillation lifetime and light yield, showing their dependence on temperature and X-ray flux as well. This approach provides a characterization method that we benchmark on a wide gamut of scintillators, including rare-earth-doped garnets and perovskite nanocrystals. Our method is particularly important for nano- and micro-scale scintillators, whose properties are challenging to quantify by conventional means: We extract the scintillation properties in perovskite nanocrystals of only a few hundreds of nanometers, observing strong photon bunching (g^((2) ) (0)>50). Our research paves the way for broader use of photon-coincidence measurement and methods from quantum optics in studying materials with complex optical properties in extremes regions of the electromagnetic spectrum.

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.

LLMs have transformed the execution of numerous tasks, including those in the medical domain. Among these, summarizing patient-reported outcomes (PROs) into concise natural language reports is of particular interest to clinicians, as it enables them to focus on critical patient concerns and spend more time in meaningful discussions. While existing work with LLMs like GPT-4 has shown impressive results, real breakthroughs could arise from leveraging SLMs as they offer the advantage of being deployable locally, ensuring patient data privacy and compliance with healthcare regulations. This study benchmarks several SLMs against LLMs for summarizing patient-reported Q\&A forms in the context of radiotherapy. Using various metrics, we evaluate their precision and reliability. The findings highlight both the promise and limitations of SLMs for high-stakes medical tasks, fostering more efficient and privacy-preserving AI-driven healthcare solutions., Comment: 10 pages

We present improvements to the pointing accuracy of the South Pole Telescope (SPT) using machine learning. The ability of the SPT to point accurately at the sky is limited by its structural imperfections, which are impacted by the extreme weather at the South Pole. Pointing accuracy is particularly important during SPT participation in observing campaigns with the Event Horizon Telescope (EHT), which requires stricter accuracy than typical observations with the SPT. We compile a training dataset of historical observations of astronomical sources made with the SPT-3G and EHT receivers on the SPT. We train two XGBoost models to learn a mapping from current weather conditions to two telescope drive control arguments -- one which corrects for errors in azimuth and the other for errors in elevation. Our trained models achieve root mean squared errors on withheld test data of $2.14''$ in cross-elevation and $3.57''$ in elevation, well below our goal of $5''$ along each axis. We deploy our models on the telescope control system and perform further in situ test observations during the EHT observing campaign in 2024 April. Our models result in significantly improved pointing accuracy: for sources within the range of input variables where the models are best trained, average combined pointing error improved 33%, from $15.9''$ to $10.6''$. These improvements, while significant, fall shy of our ultimate goal, but they serve as a proof of concept for the development of future models. Planned upgrades to the EHT receiver on the SPT will necessitate even stricter pointing accuracy which will be achievable with our methods., Comment: 23 pages, 9 figures, to be submitted to Journal of Astronomical Instrumentation (JAI)

Deployment of anti-virus software is a common strategy for preventing and controlling the propagation of computer viruses and worms over a computer network. As the deployment of such programs is often limited due to monetary or operational costs, devising optimal strategies for their allocation and deployment can be of high value to the operation, performance, and resilience of the target networks. We study the effects of anti-virus deployment (i.e., "vaccination") strategies on the ability of a network to block the spread of a virus. Such ability is obtained when the network reaches "herd immunity", achieved when a large fraction of the network entities is immune to the infection, which provides protection even for entities which are not immune. We use a model that explicitly accounts for the inherent heterogeneity of network nodes activity and derive optimal strategies for anti-virus deployment. Numerical evaluations demonstrate that the system performance is very sensitive to the chosen strategy, and thus strategies which disregard the heterogeneous spread nature may perform significantly worse relatively to those derived in this work., Comment: The extended version of the 2023 IEEE/IFIP NOMS conference paper

Recent works on accelerating Vision-Language Models show that strong performance can be maintained across a variety of vision-language tasks despite highly compressing visual information. In this work, we examine the popular acceleration approach of early pruning of visual tokens inside the language model and find that its strong performance across many tasks is not due to an exceptional ability to compress visual information, but rather the benchmarks' limited ability to assess fine-grained visual capabilities. Namely, we demonstrate a core issue with the acceleration approach where most tokens towards the top of the image are pruned away. Yet, this issue is only reflected in performance for a small subset of tasks such as localization. For the other evaluated tasks, strong performance is maintained with the flawed pruning strategy. Noting the limited visual capabilities of the studied acceleration technique, we propose FEATHER (Fast and Effective Acceleration wiTH Ensemble cRiteria), a straightforward approach that (1) resolves the identified issue with early-layer pruning, (2) incorporates uniform sampling to ensure coverage across all image regions, and (3) applies pruning in two stages to allow the criteria to become more effective at a later layer while still achieving significant speedup through early-layer pruning. With comparable computational savings, we find that FEATHER has more than $5\times$ performance improvement on the vision-centric localization benchmarks compared to the original acceleration approach., Comment: Project page: https://web.stanford.edu/~markendo/projects/feather.html

WARNING: This paper contains content that maybe upsetting or offensive to some readers. Dog whistles are coded expressions with dual meanings: one intended for the general public (outgroup) and another that conveys a specific message to an intended audience (ingroup). Often, these expressions are used to convey controversial political opinions while maintaining plausible deniability and slip by content moderation filters. Identification of dog whistles relies on curated lexicons, which have trouble keeping up to date. We introduce \textbf{FETCH!}, a task for finding novel dog whistles in massive social media corpora. We find that state-of-the-art systems fail to achieve meaningful results across three distinct social media case studies. We present \textbf{EarShot}, a novel system that combines the strengths of vector databases and Large Language Models (LLMs) to efficiently and effectively identify new dog whistles.

Leaf-cutting ants, notorious for causing defoliation in commercial forest plantations, significantly contribute to biomass and productivity losses, impacting forest producers in Brazil. These ants construct complex underground nests, highlighting the need for advanced monitoring tools to extract subsurface information across large areas. Synthetic Aperture Radar (SAR) systems provide a powerful solution for this challenge. This study presents the results of electromagnetic simulations designed to detect leaf-cutting ant nests in industrial forests. The simulations modeled nests with 6 to 100 underground chambers, offering insights into their radar signatures. Following these simulations, a field study was conducted using a drone-borne SAR operating in the P-band. A helical flight pattern was employed to generate high-resolution ground tomography of a commercial eucalyptus forest. A convolutional neural network (CNN) was implemented to detect ant nests and estimate their sizes from tomographic data, delivering remarkable results. The method achieved an ant nest detection accuracy of 100%, a false alarm rate of 0%, and an average error of 21% in size estimation. These outcomes highlight the transformative potential of integrating Synthetic Aperture Radar (SAR) systems with machine learning to enhance monitoring and management practices in commercial forestry., Comment: 8 pages, 22 figures, 4 tables

The family of finite subsets $s$ of the natural numbers such that $|s|=1+\min s$ is known as the Schreier barrier in combinatorics and Banach Space theory, and as the family of exactly $\omega$-large sets in Logic. We formulate and prove the generalizations of Friedman's Free Set and Thin Set theorems and of Rainbow Ramsey's theorem to colorings of the Schreier barrier. We analyze the strength of these theorems from the point of view of Computability Theory and Reverse Mathematics. Surprisingly, the exactly $\omega$-large counterparts of the Thin Set and Free Set theorems can code $\emptyset^{(\omega)}$, while the exactly $\omega$-large Rainbow Ramsey theorem does not code the halting set., Comment: 32 pages

Cathodoluminescence microscopy is now a well-established and powerful tool for probing the photonic properties of nanoscale materials, but in many cases, nanophotonic materials are easily damaged by the electron-beam doses necessary to achieve reasonable cathodoluminescence signal-to-noise ratios. Two-dimensional materials have proven particularly susceptible to beam-induced modifications, yielding both obstacles to high spatial-resolution measurement and opportunities for beam-induced patterning of quantum photonic systems. Here pan-sharpening techniques are applied to cathodoluminescence microscopy in order to address these challenges and experimentally demonstrate the promise of pan-sharpening for minimally-perturbative high-spatial-resolution spectrum imaging of beam-sensitive materials.

Motivated by the Iowa Fluoride Study (IFS) dataset, which comprises zero-inflated multi-level ordinal responses on tooth fluorosis, we develop an estimation scheme leveraging generalized estimating equations (GEEs) and James-Stein shrinkage. Previous analyses of this cohort study primarily focused on caries (count response) or employed a Bayesian approach to the ordinal fluorosis outcome. This study is based on the expanded dataset that now includes observations for age 23, whereas earlier works were restricted to ages 9, 13, and/or 17 according to the participants' ages at the time of measurement. The adoption of a frequentist perspective enhances the interpretability to a broader audience. Over a choice of several covariance structures, separate models are formulated for the presence (zero versus non-zero score) and severity (non-zero ordinal scores) of fluorosis, which are then integrated through shared regression parameters. This comprehensive framework effectively identifies risk or protective effects of dietary and non-dietary factors on dental fluorosis., Comment: 33 pages, 1 figure. Added Acknowledgment section and fixed page numbering

Despite the rapid integration of video perception capabilities into Large Multimodal Models (LMMs), the underlying mechanisms driving their video understanding remain poorly understood. Consequently, many design decisions in this domain are made without proper justification or analysis. The high computational cost of training and evaluating such models, coupled with limited open research, hinders the development of video-LMMs. To address this, we present a comprehensive study that helps uncover what effectively drives video understanding in LMMs. We begin by critically examining the primary contributors to the high computational requirements associated with video-LMM research and discover Scaling Consistency, wherein design and training decisions made on smaller models and datasets (up to a critical size) effectively transfer to larger models. Leveraging these insights, we explored many video-specific aspects of video-LMMs, including video sampling, architectures, data composition, training schedules, and more. For example, we demonstrated that fps sampling during training is vastly preferable to uniform frame sampling and which vision encoders are the best for video representation. Guided by these findings, we introduce Apollo, a state-of-the-art family of LMMs that achieve superior performance across different model sizes. Our models can perceive hour-long videos efficiently, with Apollo-3B outperforming most existing $7$B models with an impressive 55.1 on LongVideoBench. Apollo-7B is state-of-the-art compared to 7B LMMs with a 70.9 on MLVU, and 63.3 on Video-MME., Comment: https://apollo-lmms.github.io

Cryo-electron microscopy (cryo-EM) is an experimental technique for protein structure determination that images an ensemble of macromolecules in near-physiological contexts. While recent advances enable the reconstruction of dynamic conformations of a single biomolecular complex, current methods do not adequately model samples with mixed conformational and compositional heterogeneity. In particular, datasets containing mixtures of multiple proteins require the joint inference of structure, pose, compositional class, and conformational states for 3D reconstruction. Here, we present Hydra, an approach that models both conformational and compositional heterogeneity fully ab initio by parameterizing structures as arising from one of K neural fields. We employ a new likelihood-based loss function and demonstrate the effectiveness of our approach on synthetic datasets composed of mixtures of proteins with large degrees of conformational variability. We additionally demonstrate Hydra on an experimental dataset of a cellular lysate containing a mixture of different protein complexes. Hydra expands the expressivity of heterogeneous reconstruction methods and thus broadens the scope of cryo-EM to increasingly complex samples.

The earliest stages of star and cluster formation are hidden within dense cocoons of gas and dust, limiting their detection at optical wavelengths. With the unprecedented infrared capabilities of JWST, we can now observe dust-enshrouded star formation with $\sim$10 pc resolution out to $\sim$20 Mpc. Early findings from PHANGS-JWST suggest that 3.3 $\mu$m polycyclic aromatic hydrocarbon (PAH) emission can identify star clusters in their dust-embedded phases. Here, we extend this analysis to 19 galaxies from the PHANGS-JWST Cycle 1 Treasury Survey, providing the first characterization of compact sources exhibiting 3.3$\mu$m PAH emission across a diverse sample of nearby star-forming galaxies. We establish selection criteria, a median color threshold of F300M-F335M=0.67 at F335M=20, and identify of 1816 sources. These sources are predominantly located in dust lanes, spiral arms, rings, and galaxy centers, with $\sim$87% showing concentration indices similar to optically detected star clusters. Comparison with the PHANGS-HST catalogs suggests that PAH emission fades within $\sim$3 Myr. The H$\alpha$ equivalent width of PAH emitters is 1-2.8 times higher than that of young PHANGS-HST clusters, providing evidence that PAH emitters are on average younger. Analysis of the bright portions of luminosity functions (which should not suffer from incompleteness) shows that young dusty clusters may increase the number of optically visible $\leq$ 3 Myr-old clusters in PHANGS-HST by a factor between $\sim$1.8x-8.5x., Comment: Submitted to ApJ