Search

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 3,000 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai., Comment: 25 pages, 6 figures

Computer programming textbooks and software documentations often contain flowcharts to illustrate the flow of an algorithm or procedure. Modern OCR engines often tag these flowcharts as graphics and ignore them in further processing. In this paper, we work towards making flowchart images machine-interpretable by converting them to executable Python codes. To this end, inspired by the recent success in natural language to code generation literature, we present a novel transformer-based framework, namely FloCo-T5. Our model is well-suited for this task,as it can effectively learn semantics, structure, and patterns of programming languages, which it leverages to generate syntactically correct code. We also used a task-specific pre-training objective to pre-train FloCo-T5 using a large number of logic-preserving augmented code samples. Further, to perform a rigorous study of this problem, we introduce theFloCo dataset that contains 11,884 flowchart images and their corresponding Python codes. Our experiments show promising results, and FloCo-T5 clearly outperforms related competitive baselines on code generation metrics. We make our dataset and implementation publicly available., Comment: Published at: ICDAR 2023, Project Page: https://vl2g.github.io/projects/floco/

Tropical cyclone (TC) forecasting is crucial for disaster preparedness and mitigation. While recent deep learning approaches have shown promise, existing methods often treat TC evolution as a series of independent frame-to-frame predictions, limiting their ability to capture long-term dynamics. We present a novel application of video diffusion models for TC forecasting that explicitly models temporal dependencies through additional temporal layers. Our approach enables the model to generate multiple frames simultaneously, better capturing cyclone evolution patterns. We introduce a two-stage training strategy that significantly improves individual-frame quality and performance in low-data regimes. Experimental results show our method outperforms the previous approach of Nath et al. by 19.3% in MAE, 16.2% in PSNR, and 36.1% in SSIM. Most notably, we extend the reliable forecasting horizon from 36 to 50 hours. Through comprehensive evaluation using both traditional metrics and Fr\'echet Video Distance (FVD), we demonstrate that our approach produces more temporally coherent forecasts while maintaining competitive single-frame quality. Code accessible at https://github.com/Ren-creater/forecast-video-diffmodels., Comment: 7 pages, 7 figures

Writing comprehensive and accurate descriptions of technical drawings in patent documents is crucial to effective knowledge sharing and enabling the replication and protection of intellectual property. However, automation of this task has been largely overlooked by the research community. To this end, we introduce PatentDesc-355K, a novel large-scale dataset containing ~355K patent figures along with their brief and detailed textual descriptions extracted from more than 60K US patent documents. In addition, we propose PatentLMM - a novel multimodal large language model specifically tailored to generate high-quality descriptions of patent figures. Our proposed PatentLMM comprises two key components: (i) PatentMME, a specialized multimodal vision encoder that captures the unique structural elements of patent figures, and (ii) PatentLLaMA, a domain-adapted version of LLaMA fine-tuned on a large collection of patents. Extensive experiments demonstrate that training a vision encoder specifically designed for patent figures significantly boosts the performance, generating coherent descriptions compared to fine-tuning similar-sized off-the-shelf multimodal models. PatentDesc-355K and PatentLMM pave the way for automating the understanding of patent figures, enabling efficient knowledge sharing and faster drafting of patent documents. We make the code and data publicly available., Comment: Accepted at AAAI 2025 (Main Track). Project page: https://vl2g.github.io/projects/PatentLMM/

Let $k$ be a field of characteristic zero. Let $m$ and $\alpha$ be positive integers. For $n\geq 2$, let $R_n=k[x_1,x_2,\dots,x_n]$ with the $k$-derivation $d_n$ given by $d_n=(1-x_1x_2^{\alpha})\partial_{x_1}+x_1^m\partial_{x_2}+x_2\partial_{x_3}+\dots+x_{n-1}\partial_{x_n}$. We prove that for all odd $\alpha \geq 1$ and all odd $m\geq 3$, $d_n$ is a simple derivation on $R_n$ and $d_n(R_n)$ contains no units. This generalizes a result of D. A. Jordan. We also show that the isotropy group of $d_n$ is conjugate to a subgroup of translations., Comment: 13 pages; Comments are welcome

Indium Arsenide is a III-V semiconductor with low electron effective mass, a small band gap, strong spin-orbit coupling, and a large g-factor. These properties and its surface Fermi level pinned in the conduction band make InAs a good candidate for developing superconducting solid-state quantum devices. Here, we report the epitaxial growth of very thin InAs layers with thicknesses ranging from 12.5 nm to 500 nm grown by Molecular Beam Epitaxy on InxAl1-xAs metamorphic buffers. Differently than InAs substrates, these buffers have the advantage of being insulating at cryogenic temperatures, which allows for multiple device operations on the same wafer and thus making the approach scalable. The structural properties of the InAs layers were investigated by high-resolution X-ray diffraction, demonstrating the high crystal quality of the InAs layers. Furthermore, their transport properties, such as total and sheet carrier concentration, sheet resistance, and carrier mobility, were measured in the van der Pauw configuration at room temperature. A simple conduction model was employed to quantify the surface, bulk, and interface contributions to the overall carrier concentration and mobility.

Physics-Informed Neural Networks (PINNs) have revolutionized the computation of PDE solutions by integrating partial differential equations (PDEs) into the neural network's training process as soft constraints, becoming an important component of the scientific machine learning (SciML) ecosystem. In its current implementation, PINNs are mainly optimized using first-order methods like Adam, as well as quasi-Newton methods such as BFGS and its low-memory variant, L-BFGS. However, these optimizers often struggle with highly non-linear and non-convex loss landscapes, leading to challenges such as slow convergence, local minima entrapment, and (non)degenerate saddle points. In this study, we investigate the performance of Self-Scaled Broyden (SSBroyden) methods and other advanced quasi-Newton schemes, including BFGS and L-BFGS with different line search strategies approaches. These methods dynamically rescale updates based on historical gradient information, thus enhancing training efficiency and accuracy. We systematically compare these optimizers on key challenging linear, stiff, multi-scale and non-linear PDEs benchmarks, including the Burgers, Allen-Cahn, Kuramoto-Sivashinsky, and Ginzburg-Landau equations, and extend our study to Physics-Informed Kolmogorov-Arnold Networks (PIKANs) representation. Our findings provide insights into the effectiveness of second-order optimization strategies in improving the convergence and accurate generalization of PINNs for complex PDEs by orders of magnitude compared to the state-of-the-art., Comment: 33 pages, 27 figures

A facial morph is an image created by combining two (or more) face images pertaining to two (or more) distinct identities. Reference-free face demorphing inverts the process and tries to recover the face images constituting a facial morph without using any other information. However, there is no consensus on the evaluation metrics to be used to evaluate and compare such demorphing techniques. In this paper, we first analyze the shortcomings of the demorphing metrics currently used in the literature. We then propose a new metric called biometrically cross-weighted IQA that overcomes these issues and extensively benchmark current methods on the proposed metric to show its efficacy. Experiments on three existing demorphing methods and six datasets on two commonly used face matchers validate the efficacy of our proposed metric.

Spintronics has emerged as a field of vast applicability, and layered magnetic materials have served as a ground for advancement in this direction. Here, we report the synthesis and detailed magnetic and specific heat studies on NdSbSe, (a ZrSiS-based structure) magnetic topological material. The temperature-dependent magnetization shows the presence of competing magnetic interactions ($T_N

MXenes have attracted considerable attention in recent years due to their two-dimensional (2D) layered structure with various functionalities similar to the existing graphene and transition metal dichalcogenides. Aiming to open a new application field for MXenes in the realm of electronic devices such as ultrahigh-integrated magnetic memory, we have developed a spin-orbit torque (SOT) bilayer structure consisting of bare MXene of Cr2N: substrate//Cr2N/[Co/Pt]3/MgO using magnetron sputtering technique. We demonstrate the field-free current-induced magnetization switching (CIMS) in the Hall-cross configuration with the critical current density of ~30 MA/cm2. An attempt to replace the Cr2N layer by the Pt layer leads to the absence of field-free CIMS. As the SOT efficiency increases with increasing the Cr2N thickness, the first-principles calculation predicts the pronounced orbital-Hall conductivity in the out-of-plane component, compared to the spin-Hall conductivity in the Cr2N. X-ray magnetic circular dichroism reveals the out-of-plane uncompensated magnetic moment of Cr in the Cr2N layer at the interface, by contacting with Co in the [Co/Pt]3 ferromagnetic layer. The interfacial spin-filtering-like effect against the out-of-plane polarized spin owing to the uncompensated magnetic moment of Cr might be a possible cause for the field-free CIMS. These results show a significant milestone in the advanced 2D-spintronics, promising low-power and ultrahigh-integration.

Load forecasting is a fundamental task in smart grid. Many techniques have been applied to developing load forecasting models. Due to the challenges such as the Curse of Dimensionality, overfitting, and limited computing resources, multivariate higher-order polynomial models have received limited attention in load forecasting, despite their desirable mathematical foundations and optimization properties. In this paper, we propose low rank approximation and self-supervised dimension reduction to address the aforementioned issues. To further improve computational efficiency, we also introduce a fast Conjugate Gradient based algorithm for the proposed polynomial models. Based on the ISO New England dataset used in Global Energy Forecasting Competition 2017, the proposed method high-order polynomials with self-supervised dimension reduction (HOPS) demonstrates higher forecasting accuracy over several competitive models. Additionally, experimental results indicate that our approach alleviates redundant variable construction, achieving better forecasts with fewer input variables., Comment: 8 pages, 4 figures

We investigate the Galois module structure of the Tate-Shafarevich group of elliptic curves. For a Dirichlet character $\chi$, we give an explicit conjecture relating the ideal factorization of $L(E,\chi,1)$ to the Galois module structure of the Tate-Shafarevich group of $E/K$, where $\chi$ factors through the Galois group of $K/\mathbb{Q}$. We provide numerical evidence for this conjecture using the methods of visualization and $p$-descent. For the latter, we present a procedure that makes performing an $11$-descent over a $C_5$ number field practical for an elliptic curve $E/\mathbb{Q}$ with complex multiplication. We also expect that our method can be pushed to perform higher descents (e.g. $31$-descent) over a $C_5$ number field given more computational power., Comment: 18 pages

This work presents a first-principles study of the optoelectronic properties of vertically-stacked bilayer heterostructures composed of 2D transition-metal dichalcogenides (TMDs). The calculations are performed using the density-functional theory (DFT) and many-body perturbation theory within $G_0W_0$-BSE methodology. We aim to propose these TMD heterostructures for solar cell applications. The TMD monolayers comprising the heterojunctions considered in this work are $MoS_2$, $WS_2$, $MoSe_2$, and $WSe_2$ due to their favorable band gaps, high carrier mobility, robust absorption in the visible region, and excellent stability. These four TMD monolayers provide the basis for six heterostructures. Consequently, we have examined the structural, electronic, and optical properties of six heterostructures ($WS_2/MoS_2$, $MoSe_2/MoS_2$, $MoSe_2/WS_2$, $WSe_2/MoS_2$, $WSe_2/MoSe_2$, and $WSe_2/WS_2$). At the DFT level, all the six considered TMD heterostructures meet the essential criterion of type II band alignment, a critical factor in extending carrier lifetime. However, according to $G_0W_0$ results, $MoSe_2/WS_2$ does not exhibit the type II band alignment, instead it shows type I band alignment. The large quasiparticle gaps obtained from $G_0W_0$ approximation suggest the presence of strong electron-correlation effects. The quality of these heterojunction solar cells is estimated by computing their power conversion efficiencies (PCE). The PCEs are calculated at both the HSE06 and $G_0W_0$ levels, and the maximum PCE predicted by HSE06 calculations on our designed solar cells can reach up to 19.25% for the $WSe_2/WS_2$ heterojunction. In addition, all six TMD heterostructures are examined for their potential applications in photocatalysis for hydrogen evolution reaction, and the three of them, namely, $WS_2/MoS_2$, $MoSe_2/MoS_2$, and $WSe_2/MoS_2$ heterostructures qualify for the same., Comment: 49 pages (40 pages Manuscript and 9 pages Supporting Material), 9 figures and 11 tables in the manuscript, and 7 figures and 6 tables in the Supporting Material

We study the saturation of three-dimensional unstable perturbations on a fast rotating turbulent flow using direct numerical simulations (DNSs). Under the effect of Kolmogorov forcing, a transition between states dominated by coherent two-dimensional modes to states with three-dimensional variations (quasi-two-dimensional) is observed as we change the global rotation rate. We find this akin to a critical phenomenon, wherein the order parameter scales with the distance to the critical point raised to an exponent. The exponent itself deviates from the predicted mean field value. Also, the nature of the fluctuations of the order parameter near the critical point indicate the presence of on-off intermittency. The critical rotation rate at which the transition occurs exhibits a linear scaling behaviour with the forcing wave number. A reduced model based on linear stability analysis is used to find the linear threshold estimates; we find these to be in good agreement with the 3D nonlinear DNS results.

Variable $\gamma$-ray flares upto minute timescales reflect extreme particle acceleration sites. However, for high-redshift blazars, the detection of such rapid variations remains limited by current telescope sensitivities. Gravitationally lensed blazars serve as powerful tools to probe $\gamma$-ray production zones in distant sources, with time delays between lensed signals providing crucial insights into the spatial distribution of emission regions relative to the lens's mass-weighted center. We have utilized 15 years of Fermi-LAT $\gamma$-ray data from direction of PKS 1830$-$211 to understand the origin of flaring high-energy production zone at varying flux states. To efficiently estimate the (lensed) time delay, we used a machine learning-based tool - the Gaussian Process regression algorithm, in addition to - Autocorrelation function and Double power spectrum. We found a consistent time delay across all flaring activity states, indicating a similar location for the $\gamma$-ray emission zone, possibly within the radio core. The estimated time delay of approximately 20 days for the five flaring epochs was significantly shorter than previously estimated radio delays. This suggests that the $\gamma$-ray emission zone is closer to the central engine, in contrast to the radio emission zone, which is expected to be much farther away. A linear relationship between lag and magnification has been observed in the identified source and echo flares. Our results suggest that the $\gamma$-ray emission zone originates from similar regions away from the site of radio dissipation., Comment: 11 pages, 10 figures, 3 tables. Accepted for publication in MNRAS

Controlled modulation of electronic band structure in two-dimensional (2D) materials via doping is crucial for devices fabrication. For instance doped graphene has been envisaged for various applications like sensors, super-capacitors, transistors, p-n junctions, photo-detectors, etc. Many different techniques have been developed to achieve desired doping in 2D materials, like chemical doping, electrostatic doping, substrate doping, etc. Here, we have combined space charge doping with space and angle resolved photoemission (nano-ARPES), in order to directly observe the Fermi level modulation on micron-sized flakes of monolayer and bilayer graphene. The doping level can be tuned in a controlled manner, which allows us to directly observe the Fermi level tuning. In our experiment we successfully doped the graphene with p- and n-type carriers (holes/electrons) which are directly observed through band shift in ARPES measurements. The observed band shift is $\sim$250 meV for bilayer and $\sim$500 meV for monolayer graphene. The results from our experiment promote the space charge doping technique and nano-ARPES into other materials such as 2D semiconductors and superconductors, in order to directly observe the physical phenomena such as band gap transition and phase transition as function of carrier doping., Comment: 5 pages, 7 figures

The use of non-differentiable priors is standard in modern parsimonious Bayesian models. Lack of differentiability, however, precludes gradient-based Markov chain Monte Carlo (MCMC) methods for posterior sampling. Recently proposed proximal MCMC approaches can partially remedy this limitation. These approaches use gradients of a smooth approximation, constructed via Moreau-Yosida (MY) envelopes, to make proposals. In this work, we build an importance sampling paradigm by using the MY envelope as an importance distribution. Leveraging properties of the envelope, we establish asymptotic normality of the importance sampling estimator with an explicit expression for the asymptotic covariance matrix. Since the MY envelope density is smooth, it is amenable to gradient-based samplers. We provide sufficient conditions for geometric ergodicity of Metropolis-adjusted Langevin and Hamiltonian Monte Carlo algorithms, sampling from this importance distribution. A variety of numerical studies show that the proposed scheme can yield lower variance estimators compared to existing proximal MCMC alternatives, and is effective in both low and high dimensions., Comment: 69 pages, 11 figures

Photonics/light localization techniques are important in understanding the structural changes in biological tissues at the nano- to sub-micron scale. It is now known that structural alteration starts at the nanoscale at the beginning of cancer progression. This study examines the molecular-specific nano-structural alterations of chronic alcoholism and probiotic effects on colon cancer using a mouse model of colon cancer. Confocal microscopy and mesoscopic light-scattering analysis are applied to quantify structural changes in DNA (chromatin), cytoskeleton, and ki-67 protein cells with appropriate staining dyes. We assessed alcohol-treated and azoxymethane (AOM) with dextran sulfate sodium (DSS)-induced colitis models, including ethanol (EtOH) and probiotic (L.Casei) treatments separately and together. The inverse participation ratio (IPR) technique was employed to quantify the degree of light localization to access the molecular-specific spatial structural disorder as a biomarker for cancer progression detection. Significant enhancement of cancer progression was observed in the alcohol-treated group, and probiotics treatment with alcohol showed partial reversal of these changes in colon cancer. The results underscore the potential of the IPR technique in detecting early structural changes in colon cancer, offering insights into the mitigating effects of probiotics on alcohol-induced enhancement of colon cancer., Comment: 17 pages, 8 figures

We study a system of ultra cold dipolar Bose gas atoms confined in a two-dimensional (2D) harmonic trap with a dipolar impurity implanted at the center of the trap. Due to recent experimental progress in dipolar condensates, we focused on calculating properties of dipolar impurity systems that might guide experimentalists if they choose to study impurities in dipolar gases. We used the Gross-Pitaevskii formalism solved numerically via the split-step Crank-Nicolson method. We chose parameters of the background gas to be consistent with dysprosium (Dy), one of the strongest magnetic dipoles and of current experimental interest, and used chromium (Cr), erbium (Er), terbium (Tb), and Dy for the impurity. The dipole moments were aligned by an external field along what was chosen to be the z-axis, and studied 2D confinements that were perpendicular or parallel to the external field. We show density contour plots for the two confinements, 1D cross sections of the densities, calculated self-energies of the impurities while varying both number of atoms in the condensate and the symmetry of the trap. We also calculated the time evolution of the density of an initially pure system where an impurity is introduced. Our results found that while the self-energy increases in magnitude with increasing number of particles, it is reduced when the trap anisotropy follows the natural anisotropy of the gas, i.e., elongated along the z-axis in the case of parallel confinement. This work builds upon work done in Bose gases with zero-range interactions and demonstrates some of the features that could be found when exploring dipolar impurities in 2D Bose gases., Comment: 15 pages, 8 figures, submitted to Atoms

Human pose forecasting is inherently multimodal since multiple futures exist for an observed pose sequence. However, evaluating multimodality is challenging since the task is ill-posed. Therefore, we first propose an alternative paradigm to make the task well-posed. Next, while state-of-the-art methods predict multimodality, this requires oversampling a large volume of predictions. This raises key questions: (1) Can we capture multimodality by efficiently sampling a smaller number of predictions? (2) Subsequently, which of the predicted futures is more likely for an observed pose sequence? We address these questions with MotionMap, a simple yet effective heatmap based representation for multimodality. We extend heatmaps to represent a spatial distribution over the space of all possible motions, where different local maxima correspond to different forecasts for a given observation. MotionMap can capture a variable number of modes per observation and provide confidence measures for different modes. Further, MotionMap allows us to introduce the notion of uncertainty and controllability over the forecasted pose sequence. Finally, MotionMap captures rare modes that are non-trivial to evaluate yet critical for safety. We support our claims through multiple qualitative and quantitative experiments using popular 3D human pose datasets: Human3.6M and AMASS, highlighting the strengths and limitations of our proposed method. Project Page: https://www.epfl.ch/labs/vita/research/prediction/motionmap/, Comment: TLDR: We propose a new representation for learning multimodality in human pose forecasting which does not depend on generative models

The rise in cybercrime and the complexity of multilingual and code-mixed complaints present significant challenges for law enforcement and cybersecurity agencies. These organizations need automated, scalable methods to identify crime types, enabling efficient processing and prioritization of large complaint volumes. Manual triaging is inefficient, and traditional machine learning methods fail to capture the semantic and contextual nuances of textual cybercrime complaints. Moreover, the lack of publicly available datasets and privacy concerns hinder the research to present robust solutions. To address these challenges, we propose a framework for automated cybercrime complaint classification. The framework leverages Hinglish-adapted transformers, such as HingBERT and HingRoBERTa, to handle code-mixed inputs effectively. We employ the real-world dataset provided by Indian Cybercrime Coordination Centre (I4C) during CyberGuard AI Hackathon 2024. We employ GenAI open source model-based data augmentation method to address class imbalance. We also employ privacy-aware preprocessing to ensure compliance with ethical standards while maintaining data integrity. Our solution achieves significant performance improvements, with HingRoBERTa attaining an accuracy of 74.41% and an F1-score of 71.49%. We also develop ready-to-use tool by integrating Django REST backend with a modern frontend. The developed tool is scalable and ready for real-world deployment in platforms like the National Cyber Crime Reporting Portal. This work bridges critical gaps in cybercrime complaint management, offering a scalable, privacy-conscious, and adaptable solution for modern cybersecurity challenges.

Dynamical Ising machines achieve accelerated solving of complex combinatorial optimization problems by remapping the convergence to the ground state of the classical spin networks to the evolution of specially constructed continuous dynamical systems. The main adapted principle of constructing such systems is based on requiring that, on the one hand, the system converges to a binary state and, on the other hand, the system's energy in such states mimics the classical Ising Hamiltonian. The emergence of binary-like states is regarded to be an indispensable feature of dynamical Ising machines as it establishes the relation between the machine's continuous terminal state and the inherently discrete solution of a combinatorial optimization problem. This is emphasized by problems where the unknown quantities are represented by spin complexes, for example, the graph coloring problem. In such cases, an imprecise mapping of the continuous states to spin configurations may lead to invalid solutions requiring intensive post-processing. In contrast to such an approach, we show that there exists a class of non-binary dynamical Ising machines without the incongruity between the continuous character of the machine's states and the discreteness of the spin states. We demonstrate this feature by applying such a machine to the problems of finding proper graph coloring, constructing Latin squares, and solving Sudoku puzzles. Thus, we demonstrate that the information characterizing discrete states can be unambiguously presented in essentially continuous dynamical systems. This opens new opportunities in the realization of scalable electronic accelerators of combinatorial optimization., Comment: 21 pages

The fate of the binary neutron star (NS) merger remnants hinges sensitively upon the NS equation of state and the threshold mass, $M_{\rm ls}$, that separates a long-lived from a short-lived NS remnant. The nature of the electromagnetic counterparts is also influenced by the remnant type, particularly in determining whether a gamma-ray burst from a compact binary merger (cbGRB) is of short or long duration. We propose a novel approach to probe the threshold mass by linking it to the estimated observed ratio of long to short cbGRBs. We find that current observations broadly favour a relatively high value for this transition, $M_{\rm ls}\simeq 1.3 M_{\rm TOV}$, for which $ M_{\rm TOV} \lesssim 2.6\,M_\odot $, consistent with numerical simulations, as also shown here. Our results disfavour nuclear physics scenarios that would lead to catastrophic pressure loss at a few times nuclear density and temperatures of tens of MeV, leading to a rapid gravitational collapse of binaries with total mass $M \lesssim 1.3 M_{\rm TOV}$. Future individual gravitational wave events with on-axis cbGRBs can further bound $M_{\rm ls}$., Comment: 10 pages, 5 figures. Comments are welcome

High-quality benchmarks are the foundation for embodied AI research, enabling significant advancements in long-horizon navigation, manipulation and rearrangement tasks. However, as frontier tasks in robotics get more advanced, they require faster simulation speed, more intricate test environments, and larger demonstration datasets. To this end, we present MS-HAB, a holistic benchmark for low-level manipulation and in-home object rearrangement. First, we provide a GPU-accelerated implementation of the Home Assistant Benchmark (HAB). We support realistic low-level control and achieve over 3x the speed of previous magical grasp implementations at similar GPU memory usage. Second, we train extensive reinforcement learning (RL) and imitation learning (IL) baselines for future work to compare against. Finally, we develop a rule-based trajectory filtering system to sample specific demonstrations from our RL policies which match predefined criteria for robot behavior and safety. Combining demonstration filtering with our fast environments enables efficient, controlled data generation at scale.

Generative Adversarial Networks (GANs) have proven to be a powerful tool in generating artistic images, capable of mimicking the styles of renowned painters, such as Claude Monet. This paper introduces a tiered GAN model to progressively refine image quality through a multi-stage process, enhancing the generated images at each step. The model transforms random noise into detailed artistic representations, addressing common challenges such as instability in training, mode collapse, and output quality. This approach combines downsampling and convolutional techniques, enabling the generation of high-quality Monet-style artwork while optimizing computational efficiency. Experimental results demonstrate the architecture's ability to produce foundational artistic structures, though further refinements are necessary for achieving higher levels of realism and fidelity to Monet's style. Future work focuses on improving training methodologies and model complexity to bridge the gap between generated and true artistic images. Additionally, the limitations of traditional GANs in artistic generation are analyzed, and strategies to overcome these shortcomings are proposed.

Interpreting complex neural networks is crucial for understanding their decision-making processes, particularly in applications where transparency and accountability are essential. This proposed method addresses this need by focusing on layer-wise Relevance Propagation (LRP), a technique used in explainable artificial intelligence (XAI) to attribute neural network outputs to input features through backpropagated relevance scores. Existing LRP methods often struggle with precision in evaluating individual neuron contributions. To overcome this limitation, we present a novel approach that improves the parsing of selected neurons during LRP backward propagation, using the Visual Geometry Group 16 (VGG16) architecture as a case study. Our method creates neural network graphs to highlight critical paths and visualizes these paths with heatmaps, optimizing neuron selection through accuracy metrics like Mean Squared Error (MSE) and Symmetric Mean Absolute Percentage Error (SMAPE). Additionally, we utilize a deconvolutional visualization technique to reconstruct feature maps, offering a comprehensive view of the network's inner workings. Extensive experiments demonstrate that our approach enhances interpretability and supports the development of more transparent artificial intelligence (AI) systems for computer vision applications. This advancement has the potential to improve the trustworthiness of AI models in real-world machine vision applications, thereby increasing their reliability and effectiveness.

Object detection is a fundamental task in computer vision and image understanding, with the goal of identifying and localizing objects of interest within an image while assigning them corresponding class labels. Traditional methods, which relied on handcrafted features and shallow models, struggled with complex visual data and showed limited performance. These methods combined low-level features with contextual information and lacked the ability to capture high-level semantics. Deep learning, especially Convolutional Neural Networks (CNNs), addressed these limitations by automatically learning rich, hierarchical features directly from data. These features include both semantic and high-level representations essential for accurate object detection. This paper reviews object detection frameworks, starting with classical computer vision methods. We categorize object detection approaches into two groups: (1) classical computer vision techniques and (2) CNN-based detectors. We compare major CNN models, discussing their strengths and limitations. In conclusion, this review highlights the significant advancements in object detection through deep learning and identifies key areas for further research to improve performance.

How well can Multimodal Large Language Models (MLLMs) understand composite images? Composite images (CIs) are synthetic visuals created by merging multiple visual elements, such as charts, posters, or screenshots, rather than being captured directly by a camera. While CIs are prevalent in real-world applications, recent MLLM developments have primarily focused on interpreting natural images (NIs). Our research reveals that current MLLMs face significant challenges in accurately understanding CIs, often struggling to extract information or perform complex reasoning based on these images. We find that existing training data for CIs are mostly formatted for question-answer tasks (e.g., in datasets like ChartQA and ScienceQA), while high-quality image-caption datasets, critical for robust vision-language alignment, are only available for NIs. To bridge this gap, we introduce Composite Captions (CompCap), a flexible framework that leverages Large Language Models (LLMs) and automation tools to synthesize CIs with accurate and detailed captions. Using CompCap, we curate CompCap-118K, a dataset containing 118K image-caption pairs across six CI types. We validate the effectiveness of CompCap-118K by supervised fine-tuning MLLMs of three sizes: xGen-MM-inst.-4B and LLaVA-NeXT-Vicuna-7B/13B. Empirical results show that CompCap-118K significantly enhances MLLMs' understanding of CIs, yielding average gains of 1.7%, 2.0%, and 2.9% across eleven benchmarks, respectively.

The rapid development of Artificial Intelligence (AI) has led to the creation of powerful text generation models, such as large language models (LLMs), which are widely used for diverse applications. However, concerns surrounding AI-generated content, including issues of originality, bias, misinformation, and accountability, have become increasingly prominent. This paper offers a comprehensive overview of AI text generators (AITGs), focusing on their evolution, capabilities, and ethical implications. This paper also introduces Retrieval-Augmented Generation (RAG), a recent approach that improves the contextual relevance and accuracy of text generation by integrating dynamic information retrieval. RAG addresses key limitations of traditional models, including their reliance on static knowledge and potential inaccuracies in handling real-world data. Additionally, the paper reviews detection tools that help differentiate AI-generated text from human-written content and discusses the ethical challenges these technologies pose. The paper explores future directions for improving detection accuracy, supporting ethical AI development, and increasing accessibility. The paper contributes to a more responsible and reliable use of AI in content creation through these discussions.

Medical imaging is essential in healthcare to provide key insights into patient anatomy and pathology, aiding in diagnosis and treatment. Non-invasive techniques such as X-ray, Magnetic Resonance Imaging (MRI), Computed Tomography (CT), and Ultrasound (US), capture detailed images of organs, tissues, and abnormalities. Effective analysis of these images requires precise segmentation to delineate regions of interest (ROI), such as organs or lesions. Traditional segmentation methods, relying on manual feature-extraction, are labor-intensive and vary across experts. Recent advancements in Artificial Intelligence (AI) and Deep Learning (DL), particularly convolutional models such as U-Net and its variants (U-Net++ and U-Net 3+), have transformed medical image segmentation (MIS) by automating the process and enhancing accuracy. These models enable efficient, precise pixel-wise classification across various imaging modalities, overcoming the limitations of manual segmentation. This review explores various medical imaging techniques, examines the U-Net architectures and their adaptations, and discusses their application across different modalities. It also identifies common challenges in MIS and proposes potential solutions.

We study the preference-based pure exploration problem for bandits with vector-valued rewards. The rewards are ordered using a (given) preference cone $\mathcal{C}$ and our goal is to identify the set of Pareto optimal arms. First, to quantify the impact of preferences, we derive a novel lower bound on sample complexity for identifying the most preferred policy with a confidence level $1-\delta$. Our lower bound elicits the role played by the geometry of the preference cone and punctuates the difference in hardness compared to existing best-arm identification variants of the problem. We further explicate this geometry when the rewards follow Gaussian distributions. We then provide a convex relaxation of the lower bound and leverage it to design the Preference-based Track and Stop (PreTS) algorithm that identifies the most preferred policy. Finally, we show that the sample complexity of PreTS is asymptotically tight by deriving a new concentration inequality for vector-valued rewards.

The goal of this work is to investigate the capability of a neural operator (DeepONet) to accurately capture the complex deformation of a platelet's membrane under shear flow. The surrogate model approximated by the neural operator predicts the deformed membrane configuration based on its initial configuration and the shear stress exerted by the blood flow. The training dataset is derived from particle dynamics simulations implemented in LAMMPS. The neural operator captures the dynamics of the membrane particles with a mode error distribution of approximately 0.5\%. The proposed implementation serves as a scalable approach to integrate sub-platelet dynamics into multi-scale computational models of thrombosis.

We attempt to propose the first orthogonal-state-based protocols of measurement-device-independent quantum secure direct communication and quantum dialogue employing single basis, i.e., Bell basis as decoy qubits for eavesdropping detection. Orthogonal-state-based protocols are inherently distinct from conventional conjugate-coding protocols, offering unconditional security derived from the duality and monogamy of entanglement. Notably, these orthogonal-state-based protocols demonstrate improved performance over conjugate-coding based protocols under certain noisy environments, highlighting the significance of selecting the best basis choice of decoy qubits for secure quantum communication under collective noise. Furthermore, we rigorously analyze the security of the proposed protocols against various eavesdropping strategies, including intercept-and-resend attack, entangle-and-measure attack, information leakage attack, flip attack, and disturbance or modification attack. Our findings also show that, with appropriate modifications, the proposed orthogonal-state-based measurement-device-independent quantum secure direct communication protocol can be transformed into orthogonal-state-based measurement-device-independent versions of quantum key distribution protocols, expanding their applicability. Our protocols leverage fundamentally distinct resources to close the security loopholes linked to measurement devices, while also effectively doubling the distance for secure direct message transmission compared to traditional quantum communication methods., Comment: 13 pages, 2 figures

We study phenomena where some eigenvectors of a graph Laplacian are largely confined in small subsets of the graph. These localization phenomena are similar to those generally termed Anderson Localization in the Physics literature, and are related to the complexity of the structure of large graphs in still unexplored ways. Using spectral perturbation theory and pseudo-spectrum analysis, we explain how the presence of localized eigenvectors gives rise to fragilities (low robustness margins) to unmodeled node or link dynamics. Our analysis is demonstrated by examples of networks with relatively low complexity, but with features that appear to induce eigenvector localization. The implications of this newly-discovered fragility phenomenon are briefly discussed.

The Maxwell's equations, a system of linear partial differential equations (PDEs), describe the behavior of electric and magnetic fields in time and space and is essential for many important electromagnetic applications. Although numerical methods have been applied successfully in the past, the primary challenge in solving these equations arises from the frequency of electromagnetic fields, which depends on the shape and size of the objects to be resolved. The frequency dictates the grid size used for the spatial discretization of the PDEs. Since the domain of influence for these equations is compactly supported, even a small perturbation in frequency necessitates a new discretization of Maxwell's equations, resulting in substantial computational costs. In this work, we investigate the potential of neural operators, particularly the Deep Operator Network (DeepONet) and its variants, as a surrogate model for Maxwell's equations. Existing DeepONet implementations are restricted to real-valued data in $\mathbb{R}^n$, but since the time-harmonic Maxwell's equations yield solutions in the complex domain $\mathbb{C}^n$, a specialized architecture is required to handle complex algebra. We propose a formulation of DeepONet for complex data, define the forward pass in the complex domain, and adopt a reparametrized version of DeepONet for more efficient training. We also propose a unified framework to combine a plurality of DeepONets, trained for multiple electromagnetic field components, to incorporate the boundary condition. As shown by computational experiments, our method significantly enhances the efficiency of predicting scattered fields from a spherical object at arbitrary high frequencies.

We investigate the role of both magnetic field and chemical potential on the emergence of chaotic dynamics in the QCD confining string from the holographic principle. An earlier developed bottom-up model of Einstein-Maxwell-dilaton gravity, which mimics QCD features quite well, is used. The qualitative information about the chaos is obtained using the Poincar\'{e} sections and Lyapunov exponents. Our results depend quite strongly on the frame we consider in the analysis. In the string frame, the chemical potential and the magnetic field suppress the chaotic dynamics in both parallel and perpendicular orientations of the string with respect to the magnetic field. Meanwhile, in the Einstein frame, the magnetic field suppresses/enhances the chaotic dynamics when the string is orientated perpendicular/parallel to the magnetic field, while the chemical potential enhances the chaotic dynamics for both orientations. We further analyse the MSS bound in the parameter space of the model and find it to be always satisfied in both frames., Comment: 31 Pages, many figures, comments are welcome

Two-dimensional silicon carbide stands out among 2D materials, primarily due to its notable band gap, unlike its carbon-based counterparts. However, the binary nature and non-layered structure of bulk SiC present challenges in fabricating its 2D counterpart. Recent advancements in technology have led to the successful synthesis of atomically thin, large-scale epitaxial monolayers of hexagonal-SiC and Si9C15 , marking a significant milestone in semiconductor research. Inspired by these advancements, we have computationally designed another stable phase of 2D-SiC in the popular biphenylene network, termed SiC-biphenylene. This structure is characterized by interconnected polygons of octagons, hexagons, and tetragons arranged periodically. The dynamical and thermal stability has been confirmed through ab initio phonon dispersion and molecular dynamics simulations. The structure demonstrates a high melting point of approximately 3475 K and a direct band gap of 2.16 eV using the HSE06 functional. Upon considering many-body effects, the quasiparticle band gap widens to 2.89 eV at the G0W0 level, indicating pronounced electron correlation effects within the material. The optical spectrum obtained from solving the Bethe-Salpeter equation (G0W0+BSE) identifies the first optically active exciton peak at 2.07 eV, corresponding to a strongly bound exciton with a binding energy of 0.82 eV. Furthermore, the investigation into stable bilayer structures across various stacking configurations highlights the impact of stacking patterns on excitonic binding energies. Our investigation extends to identifying the stable bulk phase of SiC-biphenylene, revealing lower self-energy corrections compared to monolayer and bilayer structures, attributed to increased electron delocalization in bulk structures., Comment: 36 pages, 10 figures, and 89 references

Dark matter has been theorized to be charged under its own "dark electromagnetism" (dark-EM). Under this hypothesis, dark matter can behave like a cold collisionless plasma of self-interacting dark matter particles, and exhibit plasma-like instabilities with observational consequences. Using the results published in [1], which studied the degree of slowdown between two interpenetrating $e^-\,e^+$ plasma clouds due to plasma instabilities, estimates of similar interactions for colliding "dark plasmas" are explored. Comparison with astronomical observations reveals strong new constraints on dark-EM with the dark electromagnetic self-interaction $\alpha_{D} < 4 \times 10^{-25}$., Comment: 5 pages, 1 figure

Optical tweezers have revolutionized particle manipulation at the micro- and nanoscale, playing a critical role in fields such as plasmonics, biophysics, and nanotechnology. While traditional optical trapping methods primarily rely on optical forces to manipulate and organize particles, recent studies suggest that optothermal traps in surfactant solutions can induce unconventional effects such as enhanced trapping stiffness and increased diffusion. Thus, there is a need for further exploration of this system to gain a deeper understanding of the forces involved. This work investigates the behaviour of gold nanoparticles confined in an optothermal trap around a heated anchor particle in a surfactant (CTAC) solution. We observe unexpected radial confinement and synchronized rotational diffusion of particles at micrometre-scale separations from the anchor particle. These dynamics differ from known optical binding and thermophoretic effects, suggesting unexplored forces facilitated by the surfactant environment. This study expands the understanding of optothermal trapping driven by anchor plasmonic particles and introduces new possibilities for nanoparticle assembly, offering insights with potential applications in nanoscale fabrication and materials science., Comment: 24 Pages, 5 Figures

Large Language Models (LLMs) have shown to be a great success in a wide range of applications ranging from regular NLP-based use cases to AI agents. LLMs have been trained on a vast corpus of texts from various sources; despite the best efforts during the data pre-processing stage while training the LLMs, they may pick some undesirable information such as personally identifiable information (PII). Consequently, in recent times research in the area of Machine Unlearning (MUL) has become active, the main idea is to force LLMs to forget (unlearn) certain information (e.g., PII) without suffering from performance loss on regular tasks. In this work, we examine the robustness of the existing MUL techniques for their ability to enable leakage-proof forgetting in LLMs. In particular, we examine the effect of data transformation on forgetting, i.e., is an unlearned LLM able to recall forgotten information if there is a change in the format of the input? Our findings on the TOFU dataset highlight the necessity of using diverse data formats to quantify unlearning in LLMs more reliably., Comment: Accepted at EMNLP 2024 Findings; 21 pages (5 page main content + references + appendix)

Graphic designs are an effective medium for visual communication. They range from greeting cards to corporate flyers and beyond. Off-late, machine learning techniques are able to generate such designs, which accelerates the rate of content production. An automated way of evaluating their quality becomes critical. Towards this end, we introduce Design-o-meter, a data-driven methodology to quantify the goodness of graphic designs. Further, our approach can suggest modifications to these designs to improve its visual appeal. To the best of our knowledge, Design-o-meter is the first approach that scores and refines designs in a unified framework despite the inherent subjectivity and ambiguity of the setting. Our exhaustive quantitative and qualitative analysis of our approach against baselines adapted for the task (including recent Multimodal LLM-based approaches) brings out the efficacy of our methodology. We hope our work will usher more interest in this important and pragmatic problem setting., Comment: Accepted to WACV 2025. Project page: https://sahilg06.github.io/Design-o-meter/

A facial morph is an image created by combining two face images pertaining to two distinct identities. Face demorphing inverts the process and tries to recover the original images constituting a facial morph. While morph attack detection (MAD) techniques can be used to flag morph images, they do not divulge any visual information about the faces used to create them. Demorphing helps address this problem. Existing demorphing techniques are either very restrictive (assume identities during testing) or produce feeble outputs (both outputs look very similar). In this paper, we overcome these issues by proposing dc-GAN, a novel GAN-based demorphing method conditioned on the morph images. Our method overcomes morph-replication and produces high quality reconstructions of the bonafide images used to create the morphs. Moreover, our method is highly generalizable across demorphing paradigms (differential/reference-free). We conduct experiments on AMSL, FRLL-Morphs and MorDiff datasets to showcase the efficacy of our method.

We discuss the classical and quantum chaos of closed strings on a recently constructed charged confining holographic background. The confining background corresponds to the charged soliton, which is a solution of minimal $d=5$ gauged supergravity. The solution has a compact spacelike direction with a Wilson line on a circle and asymptotes to $AdS_5$ with a planar boundary. For the classical case, we analyze the chaos using the power spectrum, Poincar\'{e} sections, and Lyapunov exponents, finding that both energy and charge play constructive effects on enhancing the chaotic nature of the system. We similarly analyze quantum chaos using the distribution of the spectrum's level-spacing and out-of-time-ordered correlators and thoroughly investigate the effects of charge and energy. A gradual transition from a chaotic to an integrable regime is obtained as the energy and charge increase from lower to higher values, with charge playing a subdominant role., Comment: 28 pages, many figures. Comments are welcome

Artificial soft matter systems have appeared as important tools to harness mechanical motion for microscale manipulation. Typically, this motion is driven either by the external fields or by mutual interaction between the colloids. In the latter scenario, dynamics arise from non-reciprocal interaction among colloids within a chemical environment. In contrast, we eliminate the need for a chemical environment by utilizing a large area of optical illumination to generate thermal fields. The resulting optothermal interactions introduce non-reciprocity to the system, enabling active motion of the colloidal structure. Our approach involves two types of colloids: passive and thermally active. The thermally active colloids contain absorbing elements that capture energy from the incident optical beam, creating localized thermal fields around them. In a suspension of these colloids, the thermal gradients generated drive nearby particles through attractive thermo-osmotic forces. We investigate the resulting dynamics, which lead to various swimming modes, including active propulsion and chiral motion. We have also experimentally validated certain simulated results. By exploring the interplay between optical forces, thermal effects, and particle interactions, we aim to gain insights into controlling colloidal behavior in non-equilibrium systems. This research has significant implications for directed self-assembly, microfluidic manipulation, and the study of active matter.

This paper presents a quantum algorithm for efficiently computing partial sums and specific weighted partial sums of quantum state amplitudes. Computation of partial sums has important applications, including numerical integration, cumulative probability distributions, and probabilistic modeling. The proposed quantum algorithm uses a custom unitary construction to achieve the desired partial sums with gate complexity and circuit depth of \(O(\log_2 M)\), where \(M\) represents the number of terms in the partial sum. For cases where \(M\) is a power of two, the unitary construction is straightforward; however, for arbitrary \(M\), we develop an efficient quantum algorithm to create the required unitary matrix. Computational examples for evaluation certain partial sums and numerical integration based on our proposed algorithm are provided. We also extend the algorithm to evaluate partial sums of even or odd components and more complex weighted sums over specified intervals., Comment: 14 pages, 3 figures

We consider the problem of conditional text-to-image synthesis with diffusion models. Most recent works need to either finetune specific parts of the base diffusion model or introduce new trainable parameters, leading to deployment inflexibility due to the need for training. To address this gap in the current literature, we propose our method called TINTIN: Test-time Conditional Text-to-Image Synthesis using Diffusion Models which is a new training-free test-time only algorithm to condition text-to-image diffusion model outputs on conditioning factors such as color palettes and edge maps. In particular, we propose to interpret noise predictions during denoising as gradients of an energy-based model, leading to a flexible approach to manipulate the noise by matching predictions inferred from them to the ground truth conditioning input. This results in, to the best of our knowledge, the first approach to control model outputs with input color palettes, which we realize using a novel color distribution matching loss. We also show this test-time noise manipulation can be easily extensible to other types of conditioning, e.g., edge maps. We conduct extensive experiments using a variety of text prompts, color palettes, and edge maps and demonstrate significant improvement over the current state-of-the-art, both qualitatively and quantitatively.

Restart is a common strategy observed in nature that accelerates first-passage processes and has been extensively studied using classical random walks. In the quantum regime, restart in continuous-time quantum walks (CTQWs) has been shown to expedite the quantum hitting times. Here, we study how restarting monitored discrete-time quantum walks (DTQWs) affects the quantum hitting times. We show that the restarted DTQWs outperform classical random walks in target searches, benefiting from quantum ballistic propagation, a feature shared with their continuous-time counterparts. Moreover, the explicit coin degree of freedom in DTQWs allows them to surpass even CTQWs in target detection without sacrificing any quantum advantage. Additionally, knowledge of the target's parity or position relative to the origin can be leveraged to tailor DTQWs for even faster searches. Our study paves the way for more efficient use of DTQWs in quantum-walk-based search algorithms, simulations and modeling of quantum transport towards targeted sites in complex quantum networks., Comment: 6 pages, 8 figures

The power conversion efficiency (PCE) of organic solar cells (OSCs) has been largely improved by the introduction of novel non-fullerene acceptors (NFAs). Further improvements in PCE require a more comprehensive understanding of the free charge generation process. Recently, the small PCE of donor-acceptor blends with low offsets between the relevant frontier orbitals was attributed to inefficient exciton dissociation. However, another source of photocurrent loss is the competition between exciton diffusion and decay, which is particularly relevant for bilayers or bulk heterojunction blends with phase separated morphology. Here, we present an analytical model that combines exciton diffusion with a set of rate equations based on Marcus theory of charge transfer. An expression for the charge generation efficiency is derived from the steady-state solution of the model. Thereby, the intrinsic exciton lifetime is identified as a pivotal parameter to facilitate efficient charge generation in spite of a vanishing driving force for exciton dissociation. The dynamic formulation of the model is used to elucidate the characteristic time scales of charge generation. It is found that for low-offset systems, the pure diffusive times are considerably shorter than those associated with charge generation. It can therefore be concluded that when estimating domain sizes via exciton diffusion measurements, the assumption that excitons are instantaneously quenched at the donor-acceptor interface is only valid when a high driving force for exciton dissociation is present. The model is applied to the transient absorption dynamics of a PM6:Y6 blend. It is demonstrated that the charge generation dynamics are determined by the interplay between exciton diffusion and hole transfer kinetics, with an estimated Y6 domain size of 25nm, while interfacial charge transfer (CT) states separate rapidly into free charges.