The rapid growth of textual data across news, legal, medical, and scientific domains is becoming a challenge for efficiently accessing and understanding large volumes of content. It is increasingly complex for users to consume and extract meaningful information efficiently. Thus, raising the need for summarization. Unlike short document summarization, long document abstractive summarization is resource-intensive, and very little literature is present in this direction. BART is a widely used efficient sequence-to-sequence (seq-to-seq) model. However, when it comes to summarizing long documents, the length of the context window limits its capabilities. We proposed a model called PTS (Page-specific Target-text alignment Summarization) that extends the seq-to-seq method for abstractive summarization by dividing the source document into several pages. PTS aligns each page with the relevant part of the target summary for better supervision. Partial summaries are generated for each page of the document. We proposed another model called PTSPI (Page-specific Target-text alignment Summarization with Page Importance), an extension to PTS where an additional layer is placed before merging the partial summaries into the final summary. This layer provides dynamic page weightage and explicit supervision to focus on the most informative pages. We performed experiments on the benchmark dataset and found that PTSPI outperformed the SOTA by 6.32\% in ROUGE-1 and 8.08\% in ROUGE-2 scores.

This paper presents COT-AD, a comprehensive Dataset designed to enhance cotton crop analysis through computer vision. Comprising over 25,000 images captured throughout the cotton growth cycle, with 5,000 annotated images, COT-AD includes aerial imagery for field-scale detection and segmentation and high-resolution DSLR images documenting key diseases. The annotations cover pest and disease recognition, vegetation, and weed analysis, addressing a critical gap in cotton-specific agricultural datasets. COT-AD supports tasks such as classification, segmentation, image restoration, enhancement, deep generative model-based cotton crop synthesis, and early disease management, advancing data-driven crop management

Double perovskite-based magnets wherein frustration and competition between emergent degrees of freedom are at play can lead to novel electronic and magnetic phenomena. Herein, we report the electronic structure and magnetic properties of an ordered double perovskite material Ho2CoMnO6. In the double perovskite with general class A2BB'O6, the octahedral B and B'-site has a distinct crystallographic site. The Rietveld refinement of XRD data reveals that Ho2CoMnO6 crystallizes in the monoclinic P21/n space group. The X-ray photoelectron spectroscopy confirms the charge state of cations present in this material. The temperature dependence of magnetization and specific heat exhibit a long-range ferromagnetic ordering at Tc ~ 76 K owing to the presence of super exchange interaction between Co2+ and Mn4+ moments. Furthermore, the magnetization isotherm at 5 K shows a hysteresis curve that confirms ferromagnetic behavior of this double perovskite. We observed a re-entrant glassy state in the intermediate temperature regime, which is attributed to inherent anti-site disorder and competing interactions. A large magnetocaloric effect has been observed much below the ferromagnetic transition temperature. The temperature-dependent Raman spectroscopy studies support the presence of spin-phonon coupling and short-range order above Tc in this double perovskite. The stabilization of magnetic ordering and charge states is further analyzed through electronic structure calculations. The latter also infers the compound to be a narrow band gap insulator with the gap arising between the lower and upper Hubbard Co-d subbands. Our results demonstrate that anti-site disorder and complex 3d-4f exchange interactions in the spin-lattice account for the observed electronic and magnetic properties in this promising double perovskite material.

Adversarial training of Deep Neural Networks is known to be significantly more data-hungry when compared to standard training. Furthermore, complex data augmentations such as AutoAugment, which have led to substantial gains in standard training of image classifiers, have not been successful with Adversarial Training. We first explain this contrasting behavior by viewing augmentation during training as a problem of domain generalization, and further propose Diverse Augmentation-based Joint Adversarial Training (DAJAT) to use data augmentations effectively in adversarial training. We aim to handle the conflicting goals of enhancing the diversity of the training dataset and training with data that is close to the test distribution by using a combination of simple and complex augmentations with separate batch normalization layers during training. We further utilize the popular Jensen-Shannon divergence loss to encourage the joint learning of the diverse augmentations, thereby allowing simple augmentations to guide the learning of complex ones. Lastly, to improve the computational efficiency of the proposed method, we propose and utilize a two-step defense, Ascending Constraint Adversarial Training (ACAT), that uses an increasing epsilon schedule and weight-space smoothing to prevent gradient masking. The proposed method DAJAT achieves substantially better robustness-accuracy trade-off when compared to existing methods on the RobustBench Leaderboard on ResNet-18 and WideResNet-34-10. The code for implementing DAJAT is available here: this https URL

Quantum computing has the potential to provide exponential performance benefits in processing over classical computing. It utilizes quantum mechanics phenomena (such as superposition, entanglement, and interference) to solve a computational problem. It can explore atypical patterns over data that classical computers can't perform efficiently. Quantum computers are in the nascent stage of development and are noisy due to decoherence, i.e., quantum bits deteriorate with environmental interactions. It will take a long time for quantum computers to achieve fault tolerance although quantum algorithms can be developed in advance. Heavy investment in developing quantum hardware, software development kits, and simulators has led to multiplicity of quantum development tools. Selection of a suitable development platform requires a proper understanding of the capabilities and limitations of these tools. Although a comprehensive comparison of the different quantum development tools would be of great value, to the best of our knowledge, no such extensive study is currently available.

Datasets for training crowd counting deep networks are typically heavy-tailed in count distribution and exhibit discontinuities across the count range. As a result, the de facto statistical measures (MSE, MAE) exhibit large variance and tend to be unreliable indicators of performance across the count range. To address these concerns in a holistic manner, we revise processes at various stages of the standard crowd counting pipeline. To enable principled and balanced minibatch sampling, we propose a novel smoothed Bayesian sample stratification approach. We propose a novel cost function which can be readily incorporated into existing crowd counting deep networks to encourage strata-aware optimization. We analyze the performance of representative crowd counting approaches across standard datasets at per strata level and in aggregate. We analyze the performance of crowd counting approaches across standard datasets and demonstrate that our proposed modifications noticeably reduce error standard deviation. Our contributions represent a nuanced, statistically balanced and fine-grained characterization of performance for crowd counting approaches. Code, pretrained models and interactive visualizations can be viewed at our project page this https URL

Phase separation driven by nonequilibrium fluctuations is a hallmark of both living and synthetic active matter. Unlike equilibrium systems, where ordered states arise from the minimization of free energy, active systems are fueled by a constant injection of energy at the microscopic scale. The emergence of ordered phases in such driven systems challenges our conventional views of domain growth and interfacial structure. In this study, we investigate the coarsening of colloidal clusters in active liquids containing E. coli. Our experiments reveal that uniform dispersions of colloids and swimmers are inherently unstable, resulting in spontaneous phase separation characterized by fractal interfaces and unconventional kinetics. The correlation function of the order parameter displays dynamical scaling, with the size of colloidal domains growing as $t^{1/z}$, where $z \sim 4$, in contrast to the well-known growth laws for thermal systems with a conserved order parameter. Furthermore, the structure factor exhibits non-Porod behavior, indicating domains with fractal interfaces. This non-Porod behavior also manifests itself as a cusp singularity in the correlation function. We elucidate our experimental findings using a scalar field theory in which the nonequilibrium fluctuations arising from swimmer activity are modeled as spatio-temporally correlated noise. This coarse-grained model, which breaks time-reversal symmetry and detailed balance, successfully reproduces key experimental observations. Furthermore, it reveals a fluctuating microphase separation, where the initial domain growth, following $t^{1/4}$ scaling, is eventually arrested, thereby shedding new light on the microscopic origins of unconventional phase separation of colloids in active liquids.

Previous studies on the generalized XY model have concentrated on the equilibrium phase diagram and the equilibrium nature of distinct phases under varying parameter conditions. We direct our attention towards examining the systems evolution towards equilibrium states across different parameter values, specifically by varying the relative strengths of ferromagnetic and nematic interactions. We study the kinetics of the system, using the temporal annihilation of defects at varying temperatures and its impact on the coarsening behavior of the system. For both pure polar and pure nematic systems, we observe temperature-dependent decay of the exponent, leading to a decelerated growth of domains within the system. At parameter values where both ferromagnetic and nematic interactions are simultaneously present, we show a phase diagram highlighting three low-temperature phases : polar, nematic, and coexistence, alongside a high-temperature disordered phase. Our study provides valuable insights into the complex interplay of interactions, offering a comprehensive understanding of the systems behavior during its evolution towards equilibrium.

In this work, we present a comprehensive investigation of graphene's thermal conductivity using first-principles density functional perturbation theory calculations, with a focus on the phonon and lattice vibrational properties underlying its superior heat transport capabilities. The study highlights the role of phonon frequencies, lifetimes and mode-resolved contributions in determining graphene's thermal performance, emphasizing its high phonon group velocities and long mean free paths that contribute to thermal conductivity exceeding 3000 W/mK at room temperature. The results are compared with other two-dimensional materials like silicene (10 W/mK) and MoS2 (83 W/mK), to underline graphene's advantages in nanoscale applications. Here we report the concept of "velocity-lifetime trade-off" and use it to explain graphene's excellent invariance to high tensile and compressive strains as it exhibits minimal variation in thermal conductivity, making it an ideal material for applications requiring stability in environments with strain variability and deformation. This study establishes graphene as a benchmark material for thermal transport in next-generation 2D channel FET devices and offers a roadmap for its optimization in practical applications.

Orthogonal Time Frequency Space (OTFS) is a $\text{2-D}$ modulation technique that has the potential to overcome the challenges faced by orthogonal frequency division multiplexing (OFDM) in high Doppler environments. The performance of OTFS in a multi-user scenario with orthogonal multiple access (OMA) techniques has been impressive. Due to the requirement of massive connectivity in 5G and beyond, it is immensely essential to devise and examine the OTFS system with the existing Non-orthogonal Multiple Access (NOMA) techniques. In this paper, we propose a multi-user OTFS system based on a code-domain NOMA technique called Sparse Code Multiple Access (SCMA). This system is referred to as the OTFS-SCMA model. The framework for OTFS-SCMA is designed for both downlink and uplink. First, the sparse SCMA codewords are strategically placed on the delay-Doppler plane such that the overall overloading factor of the OTFS-SCMA system is equal to that of the underlying basic SCMA system. The receiver in downlink performs the detection in two sequential phases: first, the conventional OTFS detection using the method of linear minimum mean square error (LMMSE), and then the conventional SCMA detection. For uplink, we propose a single-phase detector based on message-passing algorithm (MPA) to detect the multiple users' symbols. The performance of the proposed OTFS-SCMA system is validated through extensive simulations both in downlink and uplink. We consider delay-Doppler planes of different parameters and various SCMA systems of overloading factor up to 200$\%$. The performance of OTFS-SCMA is compared with those of existing OTFS-OMA techniques. The comprehensive investigation demonstrates the usefulness of OTFS-SCMA in future wireless communication standards.

Many microswimmers are inherently chiral, and this chirality can introduce fascinating behaviors in a collection of microswimmers. The dynamics become even more intriguing when two types of microswimmers with distinct chirality are mixed. Our study considers a mixture of self-propelled particles with opposite chirality, examining how the system characteristics evolve as the magnitude of chirality is tuned. In weakly chiral systems, the particles exhibit similar behavior, leading to a globally flocking phase where both types of particles are well mixed. However, in an intermediate range of chirality, the particles demix and follow their trajectories, creating a competition between chirality and self-propulsion. This competition results in interesting phases within the system. We explore the characteristics of these different phases in detail, focusing on the roles of self-propulsion speed and chirality.

Learning from Demonstrations, particularly from biological experts like humans and animals, often encounters significant data acquisition challenges. While recent approaches leverage internet videos for learning, they require complex, task-specific pipelines to extract and retarget motion data for the agent. In this work, we introduce a language-model-assisted bi-level programming framework that enables a reinforcement learning agent to directly learn its reward from internet videos, bypassing dedicated data preparation. The framework includes two levels: an upper level where a vision-language model (VLM) provides feedback by comparing the learner's behavior with expert videos, and a lower level where a large language model (LLM) translates this feedback into reward updates. The VLM and LLM collaborate within this bi-level framework, using a "chain rule" approach to derive a valid search direction for reward learning. We validate the method for reward learning from YouTube videos, and the results have shown that the proposed method enables efficient reward design from expert videos of biological agents for complex behavior synthesis.

Two temperature induced phase separation(2-TIPS) is a phenomenon observed in mixtures of active and passive particles modeled by scalar activity where the temperature of the particle is proportional to its activity. The binary mixture of 'hot' and 'cold' particles phase separate when the relative temperature difference between hot and cold particles defined as activity $\chi$ exceeds a density dependent critical value. The study of kinetics in 2-TIPS, a non-equilibrium phase separation, is of fundamental importance in statistical physics. In this paper, we investigate 2-TIPS kinetics using molecular dynamics (MD) and coarse-grained (CG) modeling in 3D and 2D. The coarse-grained model couples two passive Model B equations for hot and cold particles, with coupling terms emulating the energy transfer between them by raising the temperature of cold particles and lowering that of hot particles, a key observation from the MD simulations. MD simulations reveal that at high densities, phase separation begins immediately after the quench, forming bi-continuous domains rich in hot or cold particles, similar to spinodal decomposition in passive systems. These interconnected domains are also observed in the coarse-grained model for the mixture's critical composition. Both MD and CG models show dynamic scaling of the correlation function, indicating self-similar domain growth. Regardless of dimensionality, both methods report algebraic growth in domain length with a growth exponent of $1/3$, known as the Lifshitz-Slyozov exponent, widely observed in passive systems. Our results demonstrate that the universality of phase separation kinetics observed in passive systems also extends to the non-equilibrium binary mixture undergoing 2-TIPS.

Neural Conversational QA tasks like ShARC require systems to answer questions based on the contents of a given passage. On studying recent state-of-the-art models on the ShARCQA task, we found indications that the models learn spurious clues/patterns in the dataset. Furthermore, we show that a heuristic-based program designed to exploit these patterns can have performance comparable to that of the neural models. In this paper we share our findings about four types of patterns found in the ShARC corpus and describe how neural models exploit them. Motivated by the aforementioned findings, we create and share a modified dataset that has fewer spurious patterns, consequently allowing models to learn better.

We analyse the dynamics of a weakly elastic spherical particle translating parallel to a rigid wall in a quiescent Newtonian fluid in the Stokes limit. The particle motion is constrained parallel to the wall by applying a point force and a point torque at the centre of its undeformed shape. The particle is modelled using the Navier elasticity equations. The series solutions to the Navier and the Stokes equations are utilised to obtain the displacement and velocity fields in the solid and fluid, respectively. The point force and the point torque are calculated as series in small parameters $\alpha$ and $1/H$, using the domain perturbation method and the method of reflections. Here, $\alpha$ is the measure of elastic strain induced in the particle resulting from the fluid's viscous stress, and $H$ is the non-dimensional gap width, defined as the ratio of the distance of the particle centre from the wall to its radius. The results are presented up to $\textit{O}(1/H^3)$ and $\textit{O}(1/H^2)$, assuming $\alpha \sim 1/H$, for cases where gravity is aligned and non-aligned with the particle velocity, respectively. The deformed shape of the particle is determined by the force distribution acting on it. The hydrodynamic lift due to elastic effects (acting away from the wall) appears at $\textit{O}(\alpha/H^2)$, in the former case. In an unbounded domain, the elastic effects in the latter case generate a hydrodynamic torque at \textit{O}($\alpha$) and a drag at \textit{O}($\alpha^2$). Conversely, in the former case, the torque is zero, while the drag still appears at \textit{O}($\alpha^2$).

It is known that finding approximate optima of non-convex functions is intractable. We give a simple proof to show that this problem is not even computable.

Automated feature extraction capability and significant performance of Deep Neural Networks (DNN) make them suitable for Internet of Things (IoT) applications. However, deploying DNN on edge devices becomes prohibitive due to the colossal computation, energy, and storage requirements. This paper presents a novel approach for designing and training lightweight DNN using large-size DNN. The approach considers the available storage, processing speed, and maximum allowable processing time to execute the task on edge devices. We present a knowledge distillation based training procedure to train the lightweight DNN to achieve adequate accuracy. During the training of lightweight DNN, we introduce a novel early halting technique, which preserves network resources; thus, speedups the training procedure. Finally, we present the empirically and real-world evaluations to verify the effectiveness of the proposed approach under different constraints using various edge devices.

In the recent past, a property of neural training trajectories in weight-space had been isolated, that of "local elasticity" (denoted as $S_{\rm rel}$). Local elasticity attempts to quantify the propagation of the influence of a sampled data point on the prediction at another data. In this work, we embark on a comprehensive study of the existing notion of $S_{\rm rel}$ and also propose a new definition that addresses the limitations that we point out for the original definition in the classification setting. On various state-of-the-art neural network training on SVHN, CIFAR-10 and CIFAR-100 we demonstrate how our new proposal of $S_{\rm rel}$, as opposed to the original definition, much more sharply detects the property of the weight updates preferring to make prediction changes within the same class as the sampled data. In neural regression experiments we demonstrate that the original $S_{\rm rel}$reveals a$2-$phase behavior -- that the training proceeds via an initial elastic phase when $S_{\rm rel}$ changes rapidly and an eventual inelastic phase when $S_{\rm rel}$ remains large. We show that some of these properties can be analytically reproduced in various instances of doing regression via gradient flows on model predictor classes.

This study presents a novel high-order numerical method designed for solving the two-dimensional time-fractional convection-diffusion (TFCD) equation. The Caputo definition is employed to characterize the time-fractional derivative. A weak singularity at the initial time ($t=0$) is encountered in the considered problem, which is effectively managed by adopting a discretization approach for the time-fractional derivative, where Alikhanov's high-order L2-1$_\sigma$ formula is applied on a non-uniform fitted mesh, resulting in successful tackling of the singularity. A high-order two-dimensional compact operator is implemented to approximate the spatial variables. The alternating direction implicit (ADI) approach is then employed to solve the resulting system of equations by decomposing the two-dimensional problem into two separate one-dimensional problems. The theoretical analysis, encompassing both stability and convergence aspects, has been conducted comprehensively, and it has shown that method is convergent with an order $\mathcal O\left(N_t^{-\min\{3-\alpha,\theta\alpha,1+2\alpha,2+\alpha\}}+h_x^4+h_y^4\right)$, where $\alpha\in(0,1)$ represents the order of the fractional derivative, $N_t$ is the temporal discretization parameter and $h_x$ and $h_y$ represent spatial mesh widths. Moreover, the parameter $\theta$ is utilized in the construction of the fitted mesh.