Search

Diffeomorphic deformable image registration ensures smooth invertible transformations across inspiratory and expiratory chest CT scans. Yet, in practice, deep learning-based diffeomorphic methods struggle to capture large deformations between inspiratory and expiratory volumes, and therefore lack inverse consistency. Existing methods also fail to account for model uncertainty, which can be useful for improving performance. We propose an uncertainty-aware test-time adaptation framework for inverse consistent diffeomorphic lung registration. Our method uses Monte Carlo (MC) dropout to estimate spatial uncertainty that is used to improve model performance. We train and evaluate our method for inspiratory-to-expiratory CT registration on a large cohort of 675 subjects from the COPDGene study, achieving a higher Dice similarity coefficient (DSC) between the lung boundaries (0.966) compared to both VoxelMorph (0.953) and TransMorph (0.953). Our method demonstrates consistent improvements in the inverse registration direction as well with an overall DSC of 0.966, higher than VoxelMorph (0.958) and TransMorph (0.956). Paired t-tests indicate statistically significant improvements., Comment: 5 pages, 4 figures

The impact of uniform lattice strain (with lattice parameter (LP), $a = 5.62$-$5.83$ \r{A}) and tetragonal lattice distortion (with $0.8 \leq c/a \leq1.2$ and $V_0$ volume; resulting LP: $a = 5.38$-$5.92$ \r{A}, $c = 5.33$-$6.45$ \r{A}) on the structural, electronic, and magnetic properties of CoFeCrGa SGS alloy (optimized LP: $a = 5.72$ \r{A}) has been investigated. The SGS nature of CoFeCrGa remains robust under uniform strain; however, the tetragonal lattice distortion has a detrimental impact on the SGS nature, and even a small distortions lead to metallic nature for CoFeCrGa. Despite this, the tetragonally distorted structures maintain high spin polarization (SP $\geq 60\% $); except for the structure with $c/a = 0.8$, for which SP significantly decreases to $\sim 25\%$. The lattice deformation-induced magnetic anisotropy (MA) is also explored by considering the magnetocrystalline anisotropy (MCA) and the magnetic shape anisotropy (MSA). Both MCA and MSA eliminate for the Y-I ordered and uniformly strained structures, resulting in a magnetic isotropic nature for them. Conversely, the tetragonally distorted structures with $c/a < 1.0$ and $c/a > 1.0$ exhibit very high in-plane magnetic anisotropy and perpendicular magnetic anisotropy, respectively; with magnitudes of $\sim 10^5$-$10^6$ J/m$^3$. For tetragonally deformed structures, MSA contribution to MA is negligible, and thus, MCA dictates the total MA. In summary, the isotropically strained structures show SGS nature, while the tetragonally distorted structures exhibit the high MA along with high SP.

While multi-agent systems have been shown to significantly enhance the performance of Large Language Models (LLMs) across various tasks and applications, the dense interaction between scaling agents potentially hampers their efficiency and diversity. To address these challenges, we draw inspiration from the sparse mixture-of-agents (SMoE) and propose a sparse mixture-of-agents (SMoA) framework to improve the efficiency and diversity of multi-agent LLMs. Unlike completely connected structures, SMoA introduces novel Response Selection and Early Stopping mechanisms to sparsify information flows among individual LLM agents, striking a balance between performance and efficiency. Additionally, inspired by the expert diversity principle in SMoE frameworks for workload balance between experts, we assign distinct role descriptions to each LLM agent, fostering diverse and divergent thinking. Extensive experiments on reasoning, alignment, and fairness benchmarks demonstrate that SMoA achieves performance comparable to traditional mixture-of-agents approaches but with significantly lower computational costs. Further analysis reveals that SMoA is more stable, has a greater capacity to scale, and offers considerable potential through hyper-parameter optimization. Code and data will be available at: https://github.com/David-Li0406/SMoA., Comment: Under Review

We present a review of the Semi-Symmetric Metric Gravity (SSMG) theory, representing a geometric extension of standard general relativity, based on a connection introduced by Friedmann and Schouten in 1924. The semi-symmetric connection is a connection that generalizes the Levi-Civita one, by allowing for the presence of a simple form of the torsion, described in terms of a torsion vector. The Einstein field equations are postulated to have the same form as in standard general relativity, thus relating the Einstein tensor constructed with the help of the semi-symmetric connection, with the energy-momentum tensor. The inclusion of the torsion contributions in the field equations has intriguing cosmological implications, particularly during the late-time evolution of the Universe. Presumably, these effects also dominate under high-energy conditions, and thus SSMG could potentially address unresolved issues in General Relativity and cosmology, such as the initial singularity, inflation, or the $^7$Li problem of the Big-Bang Nucleosynthesis. The explicit presence of torsion in the field equations leads to the non-conservation of the energy-momentum tensor, which can be interpreted within the irreversible thermodynamics of open systems, as describing particle creation processes. We also review in detail the cosmological applications of the theory, and investigate the statistical tests for several models, by constraining the model parameters via comparison with several observational datasets., Comment: 40 Pages, 13 Figures, Accepted for Publication in Universe

We develop methods for performing quantum teleportation of the total spin variables of an unknown state, using quantum nondemolition measurements, spin projection measurements, and classical communication. While theoretically teleportation of high-dimensional states can be attained with the assumption of generalized Bell measurements, this is typically experimentally non-trivial to implement. We introduce two protocols and show that, on average, the teleportation succeeds in teleporting the spin variables of a spin coherent state with average zero angular error in the ideal case, beating classical strategies based on quantum state estimation. In a single run of the teleportation, there is an angular error at the level of ~ 0.1 radians for large ensembles. A potential physical implementation for the scheme is with atomic ensembles and quantum nondemolition measurements performed with light. We analyze the decoherence of the protocols and find that the protocol is robust even in the limit of large ensemble sizes., Comment: 18 pages, 10 figures

The paper offers a comprehensive analysis of social networks among movie actors and directors in the film industry. Utilizing data from IMDb and Netflix, we leverage Python and NetworkX to uncover valuable insights into the movie industry's intricate web of collaborations. Key findings include identifying the top actors and directors in the OTT sector, tracking the rise of movies on OTT platforms, and analyzing centrality measures for actors. We also explore the hidden patterns within the movie data, unveiling the shortest paths between actors and predicting future collaborations. Cluster analysis categorizes movies based on various criteria, revealing the most insular and liberal clusters and identifying crossover actors bridging different segments of the industry. The study highlights that actors predominantly collaborate within language groups, transcending national boundaries. We investigate the degree of isolation of Bollywood from global cinema and identify actors working across world clusters. The project provides valuable insights into the evolving dynamics of the film industry and the impact of OTT platforms, benefiting industry professionals, scholars, and enthusiasts., Comment: 12 pages, 11 figures, 4 tables, Submitted to ICAA conference

The generation of spin-orbital currents is crucial for advancing energy-efficient spintronic devices. Here, the intricate process involved in the generation and conversion of spin and orbital to charge currents in Zr(t=2, 3, 4.5, 6, &10nm)/Co60Fe20B20(CFB), Zr/Pt/CFB, and Zr/Pt/CFB/Pt heterostructures are investigated using spin-orbital pumping ferromagnetic resonance and longitudinal spin-orbital Seebeck effect measurements. The moderate spin-orbit coupling (SOC) in the CFB layer facilitates the simultaneous generation of spin and orbital currents, which are transferred into adjacent Zr and Pt layers. Different spin-orbital to charge current contributions, namely, Inverse spin Hall effect (ISHE), Inverse orbital Hall effect (IOHE), and Inverse orbital Rashba-Edelstein effect (IOREE) are analyzed. Notably, introducing a single Pt layer increases the spin-orbital to charge current conversion via combined effects: ISHE in Pt, IOREE in Zr/Pt interface. An enhanced effective spin-orbital Hall angle ({\theta}_eff) of 0.120 {\pm} 0.004 is observed for Zr/Pt/CFB, compared to that of 0.065 {\pm} 0.002 for the Zr/CFB, and 0.077 {\pm} 0.003 for the Zr/Pt/CFB/Pt heterostructures. These findings provide new insights into orbital-moment dependent phenomena and offer promising avenues for developing advanced spintronic devices exploiting both spin and orbital degrees of freedom, even in materials with lower SOC., Comment: 21 pages, 5 figures

The variational quantum eigensolver (VQE), a type of variational quantum algorithm, is a hybrid quantum-classical algorithm to find the lowest-energy eigenstate of a particular Hamiltonian. We investigate ways to optimize the VQE solving process on multiple instances of the same problem, by observing the process on one instance of the problem to inform initialization for other processes. We aim to take advantage of the VQE solution process to obtain useful information while disregarding information which we can predict to not be very useful. In particular, we find that the solution process produces lots of data with very little new information. Therefore, we can safely disregard much of this repetitive information with little effect on the outcome of the solution process., Comment: 4 pages, 5 figures, QCCC-24 conference

Current quantum devices face challenges when dealing with large circuits due to error rates as circuit size and the number of qubits increase. The circuit wire-cutting technique addresses this issue by breaking down a large circuit into smaller, more manageable subcircuits. However, the exponential increase in the number of subcircuits and the complexity of reconstruction as more cuts are made poses a great practical challenge. Inspired by ancilla-assisted quantum process tomography and the MUBs-based grouping technique for simultaneous measurement, we propose a new approach that can reduce subcircuit running overhead. The approach first uses ancillary qubits to transform all quantum input initializations into quantum output measurements. These output measurements are then organized into commuting groups for the purpose of simultaneous measurement, based on MUBs-based grouping. This approach significantly reduces the number of necessary subcircuits as well as the total number of shots. Lastly, we provide numerical experiments to demonstrate the complexity reduction., Comment: Accepted in IEEE International Conference on Quantum Computing and Engineering - QCE24

In this paper, we investigate inaccuracy measures based on record values, focusing on the relationship between the distribution of the n-th upper and lower k-record values and the parent distribution. We extend the classical Kerridge inaccuracy measure, originally developed for comparing two distributions, to record values and derive expressions for both upper and lower record values. In addition, we explore various other inaccuracybased measures, such as cumulative residual inaccuracy, cumulative past inaccuracy, and extropy inaccuracy measures, and their applications in characterizing symmetric distributions. We compute these measures through illustrative examples for several well-known lifetime distributions, including the exponential, Pareto, and Weibull distributions. Our findings provide insights into how inaccuracy varies with record order and distribution parameters, contributing to a deeper understanding of information-theoretic measures applied to records.

We present the results of a search for gravitational-wave transients associated with core-collapse supernova SN 2023ixf, which was observed in the galaxy Messier 101 via optical emission on 2023 May 19th, during the LIGO-Virgo-KAGRA 15th Engineering Run. We define a five-day on-source window during which an accompanying gravitational-wave signal may have occurred. No gravitational waves have been identified in data when at least two gravitational-wave observatories were operating, which covered $\sim 14\%$ of this five-day window. We report the search detection efficiency for various possible gravitational-wave emission models. Considering the distance to M101 (6.7 Mpc), we derive constraints on the gravitational-wave emission mechanism of core-collapse supernovae across a broad frequency spectrum, ranging from 50 Hz to 2 kHz where we assume the GW emission occurred when coincident data are available in the on-source window. Considering an ellipsoid model for a rotating proto-neutron star, our search is sensitive to gravitational-wave energy $1 \times 10^{-5} M_{\odot} c^2$ and luminosity $4 \times 10^{-5} M_{\odot} c^2/\text{s}$ for a source emitting at 50 Hz. These constraints are around an order of magnitude more stringent than those obtained so far with gravitational-wave data. The constraint on the ellipticity of the proto-neutron star that is formed is as low as $1.04$, at frequencies above $1200$ Hz, surpassing results from SN 2019ejj., Comment: Main paper: 6 pages, 4 figures and 1 table. Total with appendices: 20 pages, 4 figures, and 1 table

Autonomy via agents using large language models (LLMs) for personalized, standardized tasks boosts human efficiency. Automating web tasks (like booking hotels within a budget) is increasingly sought after. Fulfilling practical needs, the web agent also serves as an important proof-of-concept example for various agent grounding scenarios, with its success promising advancements in many future applications. Prior research often handcrafts web agent strategies (e.g., prompting templates, multi-agent systems, search methods, etc.) and the corresponding in-context examples, which may not generalize well across all real-world scenarios. On the other hand, there has been limited study on the misalignment between a web agent's observation/action representation and the pre-training data of the LLM it's based on. This discrepancy is especially notable when LLMs are primarily trained for language completion rather than tasks involving embodied navigation actions and symbolic web elements. Our study enhances an LLM-based web agent by simply refining its observation and action space to better align with the LLM's capabilities. This approach enables our base agent to significantly outperform previous methods on a wide variety of web tasks. Specifically, on WebArena, a benchmark featuring general-purpose web interaction tasks, our agent AgentOccam surpasses the previous state-of-the-art and concurrent work by 9.8 (+29.4%) and 5.9 (+15.8%) absolute points respectively, and boosts the success rate by 26.6 points (+161%) over similar plain web agents with its observation and action space alignment. We achieve this without using in-context examples, new agent roles, online feedback or search strategies. AgentOccam's simple design highlights LLMs' impressive zero-shot performance on web tasks, and underlines the critical role of carefully tuning observation and action spaces for LLM-based agents.

Balancing safety and usefulness in large language models has become a critical challenge in recent years. Models often exhibit unsafe behavior or adopt an overly cautious approach, leading to frequent overrefusal of benign prompts, which reduces their usefulness. Addressing these issues requires methods that maintain safety while avoiding overrefusal. In this work, we examine how the overgeneration of training data using advanced teacher models (e.g., GPT-4o), including responses to both general-purpose and toxic prompts, influences the safety and overrefusal balance of instruction-following language models. Additionally, we present POROver, a strategy to use preference optimization methods in order to reduce overrefusal, via employing a superior teacher model's completions. Our results show that overgenerating completions for general-purpose prompts significantly improves the balance between safety and usefulness. Specifically, the F1 score calculated between safety and usefulness increases from 70.8% to 88.3%. Moreover, overgeneration for toxic prompts substantially reduces overrefusal, decreasing it from 94.4% to 45.2%. Furthermore, preference optimization algorithms, when applied with carefully curated preference data, can effectively reduce a model's overrefusal from 45.2% to 15.0% while maintaining comparable safety levels. Our code and data are available at https://github.com/batuhankmkaraman/POROver.

Pressure-induced emission enhancement and bandgap narrowing in vacancy-ordered halide double perovskite $Cs_2TeBr_6$ are extensively investigated through photoluminescence and absorption experiments. The below bandgap broad emission is attributed to self-trapped excitons recombination. The $Cs_2TeBr_6$ crystal, consisting of undistorted octahedra, exhibits substantial emission enhancement due to the lowering of the energy barrier between $^3P_1$ and self-trapped exciton states, as well as the suppression of nonradiative energy loss with increasing pressure. In the Raman measurements, the observed behavior of full width at half maximum of all Raman modes implies dominant electron-phonon interactions rather than anharmonic interactions between phonons. The pressure-dependent X-ray diffraction measurements reveal an anomalous behaviour in the normalized pressure as a function of the Eulerian strain.

We propose a novel scaling law for general-purpose decoder-only language models (LMs) trained on multilingual data, addressing the problem of balancing languages during multilingual pretraining. A primary challenge in studying multilingual scaling is the difficulty of analyzing individual language performance due to cross-lingual transfer. To address this, we shift the focus from individual languages to language families. We introduce and validate a hypothesis that the test cross-entropy loss for each language family is determined solely by its own sampling ratio, independent of other languages in the mixture. This insight simplifies the complexity of multilingual scaling and make the analysis scalable to an arbitrary number of languages. Building on this hypothesis, we derive a power-law relationship that links performance with dataset size, model size and sampling ratios. This relationship enables us to predict performance across various combinations of the above three quantities, and derive the optimal sampling ratios at different model scales. To demonstrate the effectiveness and accuracy of our proposed scaling law, we perform a large-scale empirical study, training more than 100 models on 23 languages spanning 5 language families. Our experiments show that the optimal sampling ratios derived from small models (85M parameters) generalize effectively to models that are several orders of magnitude larger (1.2B parameters), offering a resource-efficient approach for multilingual LM training at scale.

Efficient multi-messenger observations of gravitational-wave candidates from compact binary coalescence candidate events rely on data products reported in low-latency by the International Gravitational-wave Network (IGWN). While data products such as $\texttt{HasNS}$, the probability of at least one neutron star, and $\texttt{HasRemnant}$, the probability of remnant matter forming after merger, exist, these are not direct observables for a potential kilonova. Here, we present new kilonova light curve and ejecta mass data products derived from merger quantities measured in low latency, by marginalizing over our uncertainty in our understanding of the neutron star equation of state and using measurements of the source properties of the merger, including masses and spins. Two additional types of data products are proposed. The first is the probability of a candidate event having mass ejecta ($m_{\mathrm{ej}}$) greater than $10^{-3} M_\odot$, which we denote as $\texttt{HasEjecta}$. The second are $m_{\mathrm{ej}}$ estimates and accompanying $\texttt{ugrizy}$ and $\texttt{HJK}$ kilonova light curves predictions produced from a surrogate model trained on a grid of kilonova light curves from $\texttt{POSSIS}$, a time-dependent, three-dimensional Monte Carlo radiative transfer code. We are developing these data products in the context of the IGWN low-latency alert infrastructure, and will be advocating for their use and release for future detections.

The magnetar SGR 1935+2154 is the only known Galactic source of fast radio bursts (FRBs). FRBs from SGR 1935+2154 were first detected by CHIME/FRB and STARE2 in 2020 April, after the conclusion of the LIGO, Virgo, and KAGRA Collaborations' O3 observing run. Here we analyze four periods of gravitational wave (GW) data from the GEO600 detector coincident with four periods of FRB activity detected by CHIME/FRB, as well as X-ray glitches and X-ray bursts detected by NICER and NuSTAR close to the time of one of the FRBs. We do not detect any significant GW emission from any of the events. Instead, using a short-duration GW search (for bursts $\leq$ 1 s) we derive 50\% (90\%) upper limits of $10^{48}$ ($10^{49}$) erg for GWs at 300 Hz and $10^{49}$ ($10^{50}$) erg at 2 kHz, and constrain the GW-to-radio energy ratio to $\leq 10^{14} - 10^{16}$. We also derive upper limits from a long-duration search for bursts with durations between 1 and 10 s. These represent the strictest upper limits on concurrent GW emission from FRBs., Comment: 15 pages of text including references, 4 figures, 5 tables

As continuous learning based video analytics continue to evolve, the role of efficient edge servers in efficiently managing vast and dynamic datasets is becoming increasingly crucial. Unlike their compute architecture, storage and archival system for these edge servers has often been under-emphasized. This is unfortunate as they contribute significantly to the data management and data movement, especially in a emerging complute landscape where date storage and data protection has become one of the key concerns. To mitigate this, we propose Salient Store that specifically focuses on the integration of Computational Storage Devices (CSDs) into edge servers to enhance data processing and management, particularly in continuous learning scenarios, prevalent in fields such as autonomous driving and urban mobility. Our research, gos beyond the compute domain, and identifies the gaps in current storage system designs. We proposes a framework that aligns more closely with the growing data demands. We present a detailed analysis of data movement challenges within the archival workflows and demonstrate how the strategic integration of CSDs can significantly optimize data compression, encryption, as well as other data management tasks, to improve overall system performance. By leveraging the parallel processing capabilities of FPGAs and the high internal bandwidth of SSDs, Salient Store reduces the communication latency and data volume by ~6.2x and ~6.1x, respectively. This paper provides a comprehensive overview of the potential of CSDs to revolutionize storage, making them not just data repositories but active participants in the computational process.

Ensuring the security of released large language models (LLMs) poses a significant dilemma, as existing mechanisms either compromise ownership rights or raise data privacy concerns. To address this dilemma, we introduce TaylorMLP to protect the ownership of released LLMs and prevent their abuse. Specifically, TaylorMLP preserves the ownership of LLMs by transforming the weights of LLMs into parameters of Taylor-series. Instead of releasing the original weights, developers can release the Taylor-series parameters with users, thereby ensuring the security of LLMs. Moreover, TaylorMLP can prevent abuse of LLMs by adjusting the generation speed. It can induce low-speed token generation for the protected LLMs by increasing the terms in the Taylor-series. This intentional delay helps LLM developers prevent potential large-scale unauthorized uses of their models. Empirical experiments across five datasets and three LLM architectures demonstrate that TaylorMLP induces over 4x increase in latency, producing the tokens precisely matched with original LLMs. Subsequent defensive experiments further confirm that TaylorMLP effectively prevents users from reconstructing the weight values based on downstream datasets.

This paper introduces Gamified Adversarial Prompting (GAP), a framework that crowd-sources high-quality data for visual instruction tuning of large multimodal models. GAP transforms the data collection process into an engaging game, incentivizing players to provide fine-grained, challenging questions and answers that target gaps in the model's knowledge. Our contributions include (1) an approach to capture question-answer pairs from humans that directly address weaknesses in a model's knowledge, (2) a method for evaluating and rewarding players that successfully incentivizes them to provide high-quality submissions, and (3) a scalable, gamified platform that succeeds in collecting this data from over 50,000 participants in just a few weeks. Our implementation of GAP has significantly improved the accuracy of a small multimodal model, namely MiniCPM-Llama3-V-2.5-8B, increasing its GPT score from 0.147 to 0.477 on our dataset, approaching the benchmark set by the much larger GPT-4V. Moreover, we demonstrate that the data generated using MiniCPM-Llama3-V-2.5-8B also enhances its performance across other benchmarks, and exhibits cross-model benefits. Specifically, the same data improves the performance of QWEN2-VL-2B and QWEN2-VL-7B on the same multiple benchmarks.

We search for the galaxies associated with the intervening MgII absorbers over a redshift range of $0.4 \le z \le 1.0$ using imaging data from DESI Legacy Imaging Surveys and measure the redshift based on direct detection of nebular emission in the background quasar spectra from SDSS survey. We find 270 MgII absorbers associated with strong [O II] nebular emissions, at $2.5\sigma$ level. Among them, for 213 MgII absorbers, we detect an absorber host galaxy at impact parameters of $4 \le \rho \le 16$ kpc, including a galaxy pair associated with three absorbers, with best-fit galaxy SED model based on multi-passband photometric data from DESI Legacy Imaging surveys, supplemented with the infrared VISTA and unWISE imaging surveys. The detection rate of MgII absorber host with strong [O II] nebular emission in the finite SDSS fiber of 2--3 arcsec diameter increases from 0.2% to $\sim$3% with increasing equivalent width from $0.3$\r{A} to $\sim$ 3.5\r{A}, which remains near-constant across the probed redshift range. The associated MgII host galaxies exhibit a wide range of stellar mass from 7.94 $\le \rm log(M_{\star}/M_{\odot}) \le $ 11.11, with an average star-formation rate (SFR) of $ 5.81 \rm M_{\odot}\ yr^{-1}$. The MgII absorber hosts selected based on [O II] nebular emission mostly exhibit active star-forming systems including 13.4% starburst systems, but 10.2% with suppressed SFR. The near-constant absorption strength at low-impact parameters suggests a high gas covering fraction. We find that the MgII equivalent width ($W_{2796}$) positively correlates with SFR and specific-SFR, likely indicate their wind origin. The average velocity offset between the host and absorber suggests that the MgII gas is bound within the dark matter halo., Comment: 9 Pages, 8 Figures, submitted to A&A

Generative models can now produce photorealistic synthetic data which is virtually indistinguishable from the real data used to train it. This is a significant evolution over previous models which could produce reasonable facsimiles of the training data, but ones which could be visually distinguished from the training data by human evaluation. Recent work on OOD detection has raised doubts that generative model likelihoods are optimal OOD detectors due to issues involving likelihood misestimation, entropy in the generative process, and typicality. We speculate that generative OOD detectors also failed because their models focused on the pixels rather than the semantic content of the data, leading to failures in near-OOD cases where the pixels may be similar but the information content is significantly different. We hypothesize that estimating typical sets using self-supervised learners leads to better OOD detectors. We introduce a novel approach that leverages representation learning, and informative summary statistics based on manifold estimation, to address all of the aforementioned issues. Our method outperforms other unsupervised approaches and achieves state-of-the art performance on well-established challenging benchmarks, and new synthetic data detection tasks.

Blazars, a class of active galactic nuclei (AGN) powered by supermassive black holes, are known for their remarkable variability across multiple timescales and wavelengths. With advancements in both ground- and space-based telescopes, our understanding of AGN central engines has significantly improved. However, the mechanisms driving this variability remain elusive, and continue to fascinate both theorists and observers alike. The primary objective of this study is to constrain the X-ray variability properties of the TeV blazar PKS 2155-304. We conduct a comprehensive X-ray spectral and timing analysis, focusing on both long-term and intra-day variability. This analysis uses data from 22 epochs of XMM-Newton EPIC-pn observations, collected over 15 years (2000-2014). To investigate the variability of the source, we applied both timing and spectral analyses. For the timing analysis, we estimated fractional variability, variability amplitude, minimum variability timescales, flux distribution, and power spectral density (PSD). In the spectral analysis, we fitted the X-ray spectra using power-law, log-parabola, and broken power-law (BPL) models to determine the best-fitting parameters. Additionally, we studied the hardness ratio (HR). We observed moderate intra-day variability in most of the light curves. Seven out of the twenty-two observations showed a clear bimodal flux distribution, indicating the presence of two distinct flux states. Our analysis revealed a variable power-law PSD slope. Most HR plots did not show significant variation with flux, except for one observation (OBSID 0124930501), where HR increased with flux (Count/s). The fitted X-ray spectra favored the BPL model for the majority of observations. The findings of this work shed light on the intraday variability of blazars, providing insights into the non-thermal jet processes that drive the observed flux variations.

This paper presents a rapid approach to assessing disaster resilience in Florida, particularly regarding Hurricane Helene (2024). This category four storm made landfall on Florida's Gulf Coast in September 2024. Using the Disaster Resilience Index (DRI) developed in this paper, the preparedness and adaptive capacities of communities across counties in Florida are evaluated, identifying the most resilient areas based on three key variables: population size, average per-person income, and the Social Vulnerability Index (SVI). While the Social Vulnerability Index (SVI) accounts for factors like socioeconomic status, household composition, minority status, and housing conditions-key elements in determining a community's resilience to disasters-incorporating a county's population and per person income provides additional insight. A county's total population is directly linked to the number of individuals impacted by a disaster, while personal income reflects a household's capacity to recover. Spatial analysis was performed on the index to compare the vulnerability and resilience levels across thirty-four counties vulnerable to Hurricane Helene's projected path. The results highlight that counties with high income and lower population densities, such as Monroe and Collier, exhibit greater resilience. In contrast, areas with larger populations and higher social vulnerabilities are at greater risk of damage. This study contributes to disaster management planning by providing a rapid yet comprehensive and reassuring socioeconomic impact assessment, offering actionable insights for anticipatory measures and resource allocation.

Cerebral microbleeds, markers of brain damage from vascular and amyloid pathologies, are linked to cognitive decline in aging, but their role in Alzheimer's disease (AD) onset and progression remains unclear. This study aimed to explore whether the presence and location of lobar microbleeds are associated with amyloid-$\beta$ (A$\beta$)-PET, tau tangle formation (tau-PET), and longitudinal cognitive decline. We analyzed 1,573 ADNI participants with MR imaging data and information on the number and location of microbleeds. Associations between lobar microbleeds and pathology, cerebrospinal fluid (CSF), genetics, and cognition were examined, focusing on regional microbleeds and domain-specific cognitive decline using ordinary least-squares regression while adjusting for covariates. Cognitive decline was assessed with ADAS-Cog11 and its domain-specific sub-scores. Participants underwent neuropsychological testing at least twice, with a minimum two-year interval between assessments. Among the 1,573 participants (692 women, mean age 71.23 years), 373 participants had microbleeds. The presence of microbleeds was linked to cognitive decline, particularly in the semantic, language, and praxis domains for those with temporal lobe microbleeds. Microbleeds in the overall cortex were associated with language decline. Pathologically, temporal lobe microbleeds were associated with increased tau in the overall cortex, while cortical microbleeds were linked to elevated A$\beta$ in the temporal, parietal, and frontal regions. In this mixed population, microbleeds were connected to longitudinal cognitive decline, especially in semantic and language domains, and were associated with higher baseline A$\beta$ and tau pathology. These findings suggest that lobar microbleeds should be included in AD diagnostic and prognostic evaluations., Comment: 11 pages, 2 figures

State-of-the-art LLMs are powered by scaling -- scaling model size, dataset size and cluster size. It is economically infeasible to extensively tune hyperparameter for the largest runs. Instead, approximately optimal hyperparameters must be inferred or \textit{transferred} from smaller experiments. Hyperparameter transfer across model sizes has been studied in Yang et al. However, hyperparameter transfer across dataset size -- or token horizon -- has not been studied yet. To remedy this we conduct a large scale empirical study on how optimal learning rate (LR) depends on token horizon in LLM training. We first demonstrate that the optimal LR changes significantly with token horizon -- longer training necessitates smaller LR. Secondly we demonstrate the the optimal LR follows a scaling law, and that the optimal LR for longer horizons can be accurately estimated from shorter horizons via such scaling laws. We also provide a rule-of-thumb for transferring LR across token horizons with zero overhead over current practices. Lastly we provide evidence that LLama-1 used too high LR, and estimate the performance hit from this. We thus argue that hyperparameter transfer across data size is an important and overlooked component of LLM training.

Deep neural networks (DNNs), while increasingly deployed in many applications, struggle with robustness against anomalous and out-of-distribution (OOD) data. Current OOD benchmarks often oversimplify, focusing on single-object tasks and not fully representing complex real-world anomalies. This paper introduces a new, straightforward method employing graph structures and topological features to effectively detect both far-OOD and near-OOD data. We convert images into networks of interconnected human understandable features or visual concepts. Through extensive testing on two novel tasks, including ablation studies with large vocabularies and diverse tasks, we demonstrate the method's effectiveness. This approach enhances DNN resilience to OOD data and promises improved performance in various applications.

Large Language Models (LLMs) have revolutionized natural language processing, yet they struggle with inconsistent reasoning, particularly in novel domains and complex logical sequences. This research introduces Proof of Thought, a framework that enhances the reliability and transparency of LLM outputs. Our approach bridges LLM-generated ideas with formal logic verification, employing a custom interpreter to convert LLM outputs into First Order Logic constructs for theorem prover scrutiny. Central to our method is an intermediary JSON-based Domain-Specific Language, which by design balances precise logical structures with intuitive human concepts. This hybrid representation enables both rigorous validation and accessible human comprehension of LLM reasoning processes. Key contributions include a robust type system with sort management for enhanced logical integrity, explicit representation of rules for clear distinction between factual and inferential knowledge, and a flexible architecture that allows for easy extension to various domain-specific applications. We demonstrate Proof of Thought's effectiveness through benchmarking on StrategyQA and a novel multimodal reasoning task, showing improved performance in open-ended scenarios. By providing verifiable and interpretable results, our technique addresses critical needs for AI system accountability and sets a foundation for human-in-the-loop oversight in high-stakes domains., Comment: 38th Conference on Neural Information Processing Systems (NeurIPS 2024) System 2 Reasoning At Scale Workshop

The objective of this paper is to enhance the optimization process for neural networks by developing a dynamic learning rate algorithm that effectively integrates exponential decay and advanced anti-overfitting strategies. Our primary contribution is the establishment of a theoretical framework where we demonstrate that the optimization landscape, under the influence of our algorithm, exhibits unique stability characteristics defined by Lyapunov stability principles. Specifically, we prove that the superlevel sets of the loss function, as influenced by our adaptive learning rate, are always connected, ensuring consistent training dynamics. Furthermore, we establish the "equiconnectedness" property of these superlevel sets, which maintains uniform stability across varying training conditions and epochs. This paper contributes to the theoretical understanding of dynamic learning rate mechanisms in neural networks and also pave the way for the development of more efficient and reliable neural optimization techniques. This study intends to formalize and validate the equiconnectedness of loss function as superlevel sets in the context of neural network training, opening newer avenues for future research in adaptive machine learning algorithms. We leverage previous theoretical discoveries to propose training mechanisms that can effectively handle complex and high-dimensional data landscapes, particularly in applications requiring high precision and reliability.

The complexity of text-embedded images presents a formidable challenge in machine learning given the need for multimodal understanding of multiple aspects of expression conveyed by them. While previous research in multimodal analysis has primarily focused on singular aspects such as hate speech and its subclasses, this study expands this focus to encompass multiple aspects of linguistics: hate, targets of hate, stance, and humor. We introduce a novel dataset PrideMM comprising 5,063 text-embedded images associated with the LGBTQ+ Pride movement, thereby addressing a serious gap in existing resources. We conduct extensive experimentation on PrideMM by using unimodal and multimodal baseline methods to establish benchmarks for each task. Additionally, we propose a novel framework MemeCLIP for efficient downstream learning while preserving the knowledge of the pre-trained CLIP model. The results of our experiments show that MemeCLIP achieves superior performance compared to previously proposed frameworks on two real-world datasets. We further compare the performance of MemeCLIP and zero-shot GPT-4 on the hate classification task. Finally, we discuss the shortcomings of our model by qualitatively analyzing misclassified samples. Our code and dataset are publicly available at: https://github.com/SiddhantBikram/MemeCLIP., Comment: Accepted to EMNLP 2024 (Main)

Quantitative analysis of microscopy images is essential in the design and fabrication of components used in augmented reality/virtual reality (AR/VR) modules. However, segmenting regions of interest (ROIs) from these complex images and extracting critical dimensions (CDs) requires novel techniques, such as deep learning models which are key for actionable decisions on process, material and device optimization. In this study, we report on the fine-tuning of a pre-trained Segment Anything Model (SAM) using a diverse dataset of electron microscopy images. We employed methods such as low-rank adaptation (LoRA) to reduce training time and enhance the accuracy of ROI extraction. The model's ability to generalize to unseen images facilitates zero-shot learning and supports a CD extraction model that precisely extracts CDs from the segmented ROIs. We demonstrate the accurate extraction of binary images from cross-sectional images of surface relief gratings (SRGs) and Fresnel lenses in both single and multiclass modes. Furthermore, these binary images are used to identify transition points, aiding in the extraction of relevant CDs. The combined use of the fine-tuned segmentation model and the CD extraction model offers substantial advantages to various industrial applications by enhancing analytical capabilities, time to data and insights, and optimizing manufacturing processes.

Mixture-of-Experts (MoE) models scale more effectively than dense models due to sparse computation through expert routing, selectively activating only a small subset of expert modules. However, sparse computation challenges traditional training practices, as discrete expert routing hinders standard backpropagation and thus gradient-based optimization, which are the cornerstone of deep learning. To better pursue the scaling power of MoE, we introduce GRIN (GRadient-INformed MoE training), which incorporates sparse gradient estimation for expert routing and configures model parallelism to avoid token dropping. Applying GRIN to autoregressive language modeling, we develop a top-2 16$\times$3.8B MoE model. Our model, with only 6.6B activated parameters, outperforms a 7B dense model and matches the performance of a 14B dense model trained on the same data. Extensive evaluations across diverse tasks demonstrate the potential of GRIN to significantly enhance MoE efficacy, achieving 79.4 on MMLU, 83.7 on HellaSwag, 74.4 on HumanEval, and 58.9 on MATH., Comment: 58 pages

The current study explores the analysis of crack in initially stressed, rotating material strips, drawing insights from singular integral equations. In this work, a self-reinforced material strip with finite thickness and infinite extent, subjected to initial stress and rotational motion, has been considered to examine the Griffith fracture. The edges of the strip are pushed by constant loads from punches moving alongside it. This study makes waves in the material that affect the fracture's movement. A distinct mathematical technique is utilized to streamline the resolution of a pair of singular integral equations featuring First-order singularities. These obtained equations help us understand how the fracture behaves. The force acting at the fracture's edge is modeled using the Dirac delta function. Then, the Hilbert transformation method calculates the stress intensity factor (SIF) at the fracture's edge. Additionally, the study explores various scenarios, including constant intensity force without punch pressure, rotation parameter, initial stress, and isotropy in the strip, deduced from the SIF expression. Numerical computations and graphical analyses are conducted to assess the influence of various factors on SIF in the study. Finally, a comparison is made between the behavior of fractures in the initially stressed and rotating reinforced material strip and those in a standard material strip to identify any differences.

Both manual (relating to the use of hands) and non-manual markers (NMM), such as facial expressions or mouthing cues, are important for providing the complete meaning of phrases in American Sign Language (ASL). Efforts have been made in advancing sign language to spoken/written language understanding, but most of these have primarily focused on manual features only. In this work, using advanced neural machine translation methods, we examine and report on the extent to which facial expressions contribute to understanding sign language phrases. We present a sign language translation architecture consisting of two-stream encoders, with one encoder handling the face and the other handling the upper body (with hands). We propose a new parallel cross-attention decoding mechanism that is useful for quantifying the influence of each input modality on the output. The two streams from the encoder are directed simultaneously to different attention stacks in the decoder. Examining the properties of the parallel cross-attention weights allows us to analyze the importance of facial markers compared to body and hand features during a translating task.

Molecular optimization is a key challenge in drug discovery and material science domain, involving the design of molecules with desired properties. Existing methods focus predominantly on single-property optimization, necessitating repetitive runs to target multiple properties, which is inefficient and computationally expensive. Moreover, these methods often lack transparency, making it difficult for researchers to understand and control the optimization process. To address these issues, we propose a novel framework, Explainable Multi-property Optimization of Molecules (XMOL), to optimize multiple molecular properties simultaneously while incorporating explainability. Our approach builds on state-of-the-art geometric diffusion models, extending them to multi-property optimization through the introduction of spectral normalization and enhanced molecular constraints for stabilized training. Additionally, we integrate interpretive and explainable techniques throughout the optimization process. We evaluated XMOL on the real-world molecular datasets i.e., QM9, demonstrating its effectiveness in both single property and multiple properties optimization while offering interpretable results, paving the way for more efficient and reliable molecular design.

The ICPR 2024 Competition on Safe Segmentation of Drive Scenes in Unstructured Traffic and Adverse Weather Conditions served as a rigorous platform to evaluate and benchmark state-of-the-art semantic segmentation models under challenging conditions for autonomous driving. Over several months, participants were provided with the IDD-AW dataset, consisting of 5000 high-quality RGB-NIR image pairs, each annotated at the pixel level and captured under adverse weather conditions such as rain, fog, low light, and snow. A key aspect of the competition was the use and improvement of the Safe mean Intersection over Union (Safe mIoU) metric, designed to penalize unsafe incorrect predictions that could be overlooked by traditional mIoU. This innovative metric emphasized the importance of safety in developing autonomous driving systems. The competition showed significant advancements in the field, with participants demonstrating models that excelled in semantic segmentation and prioritized safety and robustness in unstructured and adverse conditions. The results of the competition set new benchmarks in the domain, highlighting the critical role of safety in deploying autonomous vehicles in real-world scenarios. The contributions from this competition are expected to drive further innovation in autonomous driving technology, addressing the critical challenges of operating in diverse and unpredictable environments., Comment: 15 pages, 7 figures, ICPR Competition Paper

This study employs an integral transform approach for Love wave propagation in a rotating composite structure having an interfacial crack. The structure comprises an initially stressed functionally graded piezoelectric viscoelastic half-space bonded to a piezoelectric viscoelastic half-space. The study focuses on two material systems: Epoxy-BNKLBT paired with Epoxy-KNLNTS and Epoxy-BNKLBT paired with Epoxy-PZT7A. The viscoelastic materials are modeled to reflect their complex behavior under rotational and stress conditions. The Galilean transformation is applied to convert the Cartesian coordinates system into a moving reference frame aligned with the Love wave's propagation. Employing Bessel function properties, the system is converted into a set of double integral equations and subsequently reformulated into simultaneous Fredholm integral equations. Numerical solutions to these Fredholm integral equations are used to calculate the electric displacement intensity factor (EDIF) and stress intensity factor (SIF) near the interfacial crack. The key objective of this study is to visualize the impact of different material parameters, like piezoelectric constants, dielectric constants, initial stress, interface electric displacement, interface stress, and rotation, on SIF and EDIF. The investigations of this study will be helpful for advanced technologies like surface acoustic wave (SAW) sensors and piezoelectric actuators, as well as to enhance SAW bio-sensor sensitivity and stability for early cancer detection and biomedical implants.

A popular new method in mechanistic interpretability is to train high-dimensional sparse autoencoders (SAEs) on neuron activations and use SAE features as the atomic units of analysis. However, the body of evidence on whether SAE feature spaces are useful for causal analysis is underdeveloped. In this work, we use the RAVEL benchmark to evaluate whether SAEs trained on hidden representations of GPT-2 small have sets of features that separately mediate knowledge of which country a city is in and which continent it is in. We evaluate four open-source SAEs for GPT-2 small against each other, with neurons serving as a baseline, and linear features learned via distributed alignment search (DAS) serving as a skyline. For each, we learn a binary mask to select features that will be patched to change the country of a city without changing the continent, or vice versa. Our results show that SAEs struggle to reach the neuron baseline, and none come close to the DAS skyline. We release code here: https://github.com/MaheepChaudhary/SAE-Ravel

We report on the first experimental characterization of a laser-wakefield accelerator able to deliver, in a single pulse, doses in excess of 1 Gy on timescales of the order of tens of femtoseconds, reaching unprecedented average dose-rates above $10^{13}$ Gy/s. The irradiator is demonstrated to deliver doses tuneable up to 2.2 Gy in a cm$^2$ area and with a high degree of longitudinal and transverse uniformity in a single irradiation. In this regime, proof-of-principle irradiation of patient-derived glioblastoma stem-like cells and human skin fibroblast cells show indications of a differential cellular response, when compared to reference irradiations at conventional dose-rates. These include a statistically significant increase in relative biological effectiveness ($1.40\pm0.08$ at 50\% survival for both cell lines) and a significant reduction of the relative radioresistance of tumour cells. Data analysis provides preliminary indications that these effects might not be fully explained by induced oxygen depletion in the cells but may be instead linked to a higher complexity of the damages triggered by the ultra-high density of ionising tracks of femtosecond-scale radiation pulses. These results demonstrate an integrated platform for systematic radiobiological studies at unprecedented beam durations and dose-rates, a unique infrastructure for translational research in radiobiology at the femtosecond scale., Comment: 29 pages, 11 figures, Paper submission

This study explores the impact of upselling on user engagement. We model users' deposit behaviour on the fantasy sports platform Dream11. Subsequently, we develop an experimental framework to evaluate the effect of upselling using an intensity parameter. Our live experiments on user deposit behaviour reveal decreased user recall with heightened upselling intensity. Our findings indicate that increased upselling intensity improves user deposit metrics and concurrently diminishes user satisfaction and conversion rates. We conduct robust counterfactual analysis and train causal meta-learners to personalise users' upselling intensity levels to reach an optimal trade-off point.

AI Infrastructure plays a key role in the speed and cost-competitiveness of developing and deploying advanced AI models. The current demand for powerful AI infrastructure for model training is driven by the emergence of generative AI and foundational models, where on occasion thousands of GPUs must cooperate on a single training job for the model to be trained in a reasonable time. Delivering efficient and high-performing AI training requires an end-to-end solution that combines hardware, software and holistic telemetry to cater for multiple types of AI workloads. In this report, we describe IBM's hybrid cloud infrastructure that powers our generative AI model development. This infrastructure includes (1) Vela: an AI-optimized supercomputing capability directly integrated into the IBM Cloud, delivering scalable, dynamic, multi-tenant and geographically distributed infrastructure for large-scale model training and other AI workflow steps and (2) Blue Vela: a large-scale, purpose-built, on-premises hosting environment that is optimized to support our largest and most ambitious AI model training tasks. Vela provides IBM with the dual benefit of high performance for internal use along with the flexibility to adapt to an evolving commercial landscape. Blue Vela provides us with the benefits of rapid development of our largest and most ambitious models, as well as future-proofing against the evolving model landscape in the industry. Taken together, they provide IBM with the ability to rapidly innovate in the development of both AI models and commercial offerings., Comment: Corresponding Authors: Talia Gershon, Seetharami Seelam,Brian Belgodere, Milton Bonilla

This paper deals with time-fractional stochastic Navier-Stokes equations, which are characterized by the coexistence of stochastic noise and a fractional power of the Laplacian. We establish sufficient conditions for the existence and approximate controllability of a unique mild solution to time-fractional stochastic Navier-Stokes equations. Using a fixed point technique, we first demonstrate the existence and uniqueness of a mild solution to the equation under consideration. We then establish approximate controllability results by using the concepts of fractional calculus, semigroup theory, functional analysis and stochastic analysis.

In this study, we achieved the improvement of uniaxial perpendicular magnetic anisotropy (PMA) in the W/MgAl2O4/CoFeMnSi/W/CoFeMnSi/MgAl2O4/W heterostructure by manipulating the annealing temperature (TA) [350 C, 450 C, and 550 C]. We observed a maximum effective PMA energy density (Keff) of = 1.604 x 106 erg/cc with low saturation magnetization (Ms) at the specified TA. The enhancement of Keff with Ms is significantly influenced by structural variations at the interfaces of CoFeMnSi and MgAl2O4, attributed to sufficient interfacial oxidation dependent on the TA. The TA was identified as a critical factor affecting the surface morphology, grain size, and surface roughness of the multilayer. Fourier-transform infrared (FT-IR) measurements were employed to confirm the presence of Co-O or Fe-O bond in the multilayer structures, elucidating the true origin of PMA. The control of interfacial oxidation at the interface during annealing is crucial for regulating the strength of PMA. Therefore, this double CoFeMnSi/MgAl2O4-based multilayer presents a promising avenue, serving as a favorable candidate for future p-MTJs-based spintronic devices with enhanced thermal stability.

The deployment of Deep Neural Networks in energy-constrained environments, such as Energy Harvesting Wireless Sensor Networks, presents unique challenges, primarily due to the intermittent nature of power availability. To address these challenges, this study introduces and evaluates a novel training methodology tailored for DNNs operating within such contexts. In particular, we propose a dynamic dropout technique that adapts to both the architecture of the device and the variability in energy availability inherent in energy harvesting scenarios. Our proposed approach leverages a device model that incorporates specific parameters of the network architecture and the energy harvesting profile to optimize dropout rates dynamically during the training phase. By modulating the network's training process based on predicted energy availability, our method not only conserves energy but also ensures sustained learning and inference capabilities under power constraints. Our preliminary results demonstrate that this strategy provides 6 to 22 percent accuracy improvements compared to the state of the art with less than 5 percent additional compute. This paper details the development of the device model, describes the integration of energy profiles with intermittency aware dropout and quantization algorithms, and presents a comprehensive evaluation of the proposed approach using real-world energy harvesting data.