Search

ANCHOLIK-NER is a linguistically diverse dataset for Named Entity Recognition (NER) in Bangla regional dialects, capturing variations across Sylhet, Chittagong, and Barishal. The dataset has around 10,443 sentences, 3,481 sentences per region. The data was collected from two publicly available datasets and through web scraping from various online newspapers, articles. To ensure high-quality annotations, the BIO tagging scheme was employed, and professional annotators with expertise in regional dialects carried out the labeling process. The dataset is structured into separate subsets for each region and is available both in CSV format. Each entry contains textual data along with identified named entities and their corresponding annotations. Named entities are categorized into ten distinct classes: Person, Location, Organization, Food, Animal, Colour, Role, Relation, Object, and Miscellaneous. This dataset serves as a valuable resource for developing and evaluating NER models for Bangla dialectal variations, contributing to regional language processing and low-resource NLP applications. It can be utilized to enhance NER systems in Bangla dialects, improve regional language understanding, and support applications in machine translation, information retrieval, and conversational AI.

The DNS (Domain Name System) protocol has been in use since the early days of the Internet. Although DNS as a de facto networking protocol had no security considerations in its early years, there have been many security enhancements, such as DNSSec (Domain Name System Security Extensions), DoT (DNS over Transport Layer Security), DoH (DNS over HTTPS) and DoQ (DNS over QUIC). With all these security improvements, it is not yet clear what resource-constrained Internet-of-Things (IoT) devices should be used for robustness. In this paper, we investigate different DNS security approaches using an edge DNS resolver implemented as a Virtual Network Function (VNF) to replicate the impact of the protocol from an IoT perspective and compare their performances under different conditions. We present our results for cache-based and non-cached responses and evaluate the corresponding security benefits. Our results and framework can greatly help consumers, manufacturers, and the research community decide and implement their DNS protocols depending on the given dynamic network conditions and enable robust Internet access via DNS for different devices., Comment: 6 pages, 2 tables, 2 figures. This paper has been accepted in the 2025 IEEE International Conference on Communications (ICC): SAC Cloud Computing, Networking, and Storage Track. The final version will be published in the IEEE Xplore

Aligning visual features with language embeddings is a key challenge in vision-language models (VLMs). The performance of such models hinges on having a good connector that maps visual features generated by a vision encoder to a shared embedding space with the LLM while preserving semantic similarity. Existing connectors, such as multilayer perceptrons (MLPs), often produce out-of-distribution or noisy inputs, leading to misalignment between the modalities. In this work, we propose a novel vision-text alignment method, AlignVLM, that maps visual features to a weighted average of LLM text embeddings. Our approach leverages the linguistic priors encoded by the LLM to ensure that visual features are mapped to regions of the space that the LLM can effectively interpret. AlignVLM is particularly effective for document understanding tasks, where scanned document images must be accurately mapped to their textual content. Our extensive experiments show that AlignVLM achieves state-of-the-art performance compared to prior alignment methods. We provide further analysis demonstrating improved vision-text feature alignment and robustness to noise.

This report presents our approach for the IEEE SP Cup 2025: Deepfake Face Detection in the Wild (DFWild-Cup), focusing on detecting deepfakes across diverse datasets. Our methodology employs advanced backbone models, including MaxViT, CoAtNet, and EVA-02, fine-tuned using supervised contrastive loss to enhance feature separation. These models were specifically chosen for their complementary strengths. Integration of convolution layers and strided attention in MaxViT is well-suited for detecting local features. In contrast, hybrid use of convolution and attention mechanisms in CoAtNet effectively captures multi-scale features. Robust pretraining with masked image modeling of EVA-02 excels at capturing global features. After training, we freeze the parameters of these models and train the classification heads. Finally, a majority voting ensemble is employed to combine the predictions from these models, improving robustness and generalization to unseen scenarios. The proposed system addresses the challenges of detecting deepfakes in real-world conditions and achieves a commendable accuracy of 95.83% on the validation dataset., Comment: Technical report for IEEE Signal Processing Cup 2025, 7 pages

This article reports the development of a versatile high-speed reciprocating drive system (HRDS) with interchangeable probe heads to characterize the edge plasma region of ADITYA-U tokamak. This reciprocating probe drive system consisting of Langmuir and magnetic probe heads, is designed, fabricated, installed, and operated for studying the extent of fuel/impurity gas propagation and its influence on plasma dynamics in the far-edge region inside the last closed magnetic flux surface (LCFS). The HRDS is driven by a highly accurate, easy-to-control, dynamic, brushless, permanently excited synchronous servo motor operated by a PXI-commanded controller. The system is remotely operated and allows for precise control of the speed, acceleration, and distance traveled of the probe head on a shot-to-shot basis, facilitating seamless control of operations according to experimental requirements. Using this system, consisting of a linear array of Langmuir probes, measurements of plasma density, temperature, potential, and their fluctuations revealed that the fuel gas-puff impact these mean and fluctuating parameters up to three to four cm inside the LCFS. Attaching an array of magnetic probes to this system led to measurements of magnetic fluctuations inside the LCFS. The HRDS system is fully operational and serves as an important diagnostic tool for ADITYA-U tokamak.

Experiments on ADITYA-U tokamak show a marked enhancement in the sawtooth period by application of short gas puffs of fuel that cause a modification of the radial density profile. A consequent suppression of the trapped electron modes (TEMs) then leads to an increase in the core electron temperature. This slows down the heat propagation following a sawtooth crash, causing a delay in achieving the critical temperature gradient inside the q = 1 surface required for the next sawtooth crash to happen. The overall scenario has strong similarities with the behavior of sawtooth under electron cyclotron resonance heating (ECRH). Our findings suggest an alternate, simpler technique for sawtooth control that may be usefully employed in small/medium-sized tokamaks that do not have an ECRH or any other auxiliary heating facility.

Wireless local area networks remain vulnerable to attacks initiated during the connection establishment (CE) phase. Current Wi-Fi security protocols fail to fully mitigate attacks like man-in-the-middle, preamble spoofing, and relaying. To fortify the CE phase, in this paper we design a backward-compatible scheme using a digital signature interwoven into the preambles at the physical (PHY) layer with time constraints to effectively counter those attacks. This approach slices a MAC-layer signature and embeds the slices within CE frame preambles without extending frame size, allowing one or multiple stations to concurrently verify their respective APs' transmissions. The concurrent CEs are supported by enabling the stations to analyze the consistent patterns of PHY-layer headers and identify whether the received frames are the anticipated ones from the expected APs, achieving 100% accuracy without needing to examine their MAC-layer headers. Additionally, we design and implement a fast relay attack to challenge our proposed defense and determine its effectiveness. We extend existing open-source tools to support IEEE 802.11ax to evaluate the effectiveness and practicality of our proposed scheme in a testbed consisting of USRPs, commercial APs, and Wi-Fi devices, and we show that our relay attack detection achieves 96-100% true positive rates. Finally, end-to-end formal security analyses confirm the security and correctness of the proposed solution., Comment: 13 pages

Star clusters, including high-mass stars, form within hub-filament systems (HFSs). Observations of HFSs that remain unaffected by feedback from embedded stars are rare yet crucial for understanding the mass inflow process in high-mass star formation. Using the JWST NIRCAM images, Dewangan et al. 2024, reported that the high-mass protostar G11P1 is embedded in a candidate HFS (G11P1-HFS; $<0.6$ pc). Utilizing ALMA N$_{2}$H$^{+}$(1-0) data, we confirm the presence of G11P1-HFS and study the dense gas kinematics. We analyzed the position-position-velocity (PPV) map and estimated on-sky velocity gradient ($V_g$) and gravity ($\mathcal{F}_{g}$) vectors. The spatial distribution of gas velocity and H$_2$ column density was examined. The steep $V_g$ of 5 km s$^{-1}$ pc$^{-1}$ and $-$7 km s$^{-1}$ pc$^{-1}$ toward either side of G11P1-hub, and the decreasing $V_g$ toward the hub, identify G11P1-HFS as a small-scale HFS in its nascent phase. $V_g$ and $\mathcal{F}_{g}$ align along the filaments, indicating gravity-driven flows. This work highlights the wiggled, funnel-shaped morphology of a HFS in PPV space, suggesting the importance of subfilaments or transverse gas flows in mass transportation to the hub., Comment: 10 pages, 11 figures, 1 table, Accepted for publication in A&A Letters

We construct parameterized families of imaginary (resp. real) quadratic fields whose class groups have $n$-rank at least $2$., Comment: 7 pages. Published in `Class Groups of Number Fields and Related Topics', Springer Proceedings in Mathematics & Statistics, vol 470, Springer, Singapore

Teleoperation for robot imitation learning is bottlenecked by hardware availability. Can high-quality robot data be collected without a physical robot? We present a system for augmenting Apple Vision Pro with real-time virtual robot feedback. By providing users with an intuitive understanding of how their actions translate to robot motions, we enable the collection of natural barehanded human data that is compatible with the limitations of physical robot hardware. We conducted a user study with 15 participants demonstrating 3 different tasks each under 3 different feedback conditions and directly replayed the collected trajectories on physical robot hardware. Results suggest live robot feedback dramatically improves the quality of the collected data, suggesting a new avenue for scalable human data collection without access to robot hardware. Videos and more are available at https://nataliya.dev/armada.

We present HadaCore, a modified Fast Walsh-Hadamard Transform (FWHT) algorithm optimized for the Tensor Cores present in modern GPU hardware. HadaCore follows the recursive structure of the original FWHT algorithm, achieving the same asymptotic runtime complexity but leveraging a hardware-aware work decomposition that benefits from Tensor Core acceleration. This reduces bottlenecks from compute and data exchange. On Nvidia A100 and H100 GPUs, HadaCore achieves speedups of 1.1-1.4x and 1.0-1.3x, with a peak gain of 3.5x and 3.6x respectively, when compared to the existing state-of-the-art implementation of the original algorithm. We also show that when using FP16 or BF16, our implementation is numerically accurate, enabling comparable accuracy on MMLU benchmarks when used in an end-to-end Llama3 inference run with quantized (FP8) attention.

The Alternate Wetting and Drying (AWD) method is a rice-growing water management technique promoted as a sustainable alternative to Continuous Flooding (CF). Climate change has placed the agricultural sector in a challenging position, particularly as global water resources become increasingly scarce, affecting rice production on irrigated lowlands. Rice, a staple food for over half of the world's population, demands significantly more water than other major crops. In Bangladesh, Boro rice, in particular, requires considerable water inputs during its cultivation. Traditionally, farmers manually measure water levels, a process that is both time-consuming and prone to errors. While ultrasonic sensors offer improvements in water height measurement, they still face limitations, such as susceptibility to weather conditions and environmental factors. To address these issues, we propose a novel approach that automates water height measurement using computer vision, specifically through a convolutional neural network (CNN). Our attention-based architecture achieved an $R^2$ score of 0.9885 and a Mean Squared Error (MSE) of 0.2766, providing a more accurate and efficient solution for managing AWD systems., Comment: 15 pages, 6 figures

The sensitivity of Global Navigation Satellite Systems (GNSS) receivers to ionospheric disturbances and their constant growth are nowadays resulting in an increased concern of GNSS-users about the impacts of ionospheric disturbances at mid-latitudes. The geomagnetic storm of June 2015 is an example of a rare phenomenon of a spill-over of equatorial plasma bubbles well North from their habitual. We study the occurrence of small- and medium-scale irregularities in the North Atlantic Eastern-Mediterranean mid- and low-latitudinal zone by analysing the behaviour of the amplitude scintillation index S4 and of the rate of total electron content index (ROTI) during such a storm. In addition, large scale perturbations of the ionospheric electron density were studied using ground and space-born instruments, thus characterizing a complex perturbation behaviour over the region mentioned above. The multi-source data allows us to characterize the impact of irregularities of different scales to better understand the ionospheric dynamics and stress the importance of a proper monitoring of the ionosphere in the studied region.

With the approach of the High Luminosity Large Hadron Collider (HL-LHC) era set to begin particle collisions by the end of this decade, it is evident that the computational demands of traditional collision simulation methods are becoming increasingly unsustainable. Existing approaches, which rely heavily on first-principles Monte Carlo simulations for modeling event showers in calorimeters, are projected to require millions of CPU-years annually -- far exceeding current computational capacities. This bottleneck presents an exciting opportunity for advancements in computational physics by integrating deep generative models with quantum simulations. We propose a quantum-assisted hierarchical deep generative surrogate founded on a variational autoencoder (VAE) in combination with an energy conditioned restricted Boltzmann machine (RBM) embedded in the model's latent space as a prior. By mapping the topology of D-Wave's Zephyr quantum annealer (QA) into the nodes and couplings of a 4-partite RBM, we leverage quantum simulation to accelerate our shower generation times significantly. To evaluate our framework, we use Dataset 2 of the CaloChallenge 2022. Through the integration of classical computation and quantum simulation, this hybrid framework paves way for utilizing large-scale quantum simulations as priors in deep generative models., Comment: Neurips ML4PS 2024. 5 Figs, 8 pp

Let $C_\ell/\mathbb Q$ denote the curve with affine model $y^\ell=x(x^\ell-1)$, where $\ell\geq 3$ is prime. In this paper we study the limiting distributions of the normalized $L$-polynomials of the curves by computing their Sato-Tate groups and distributions. We also provide results for the number of points on the curves over finite fields, including a formula in terms of Jacobi sums when the field $\mathbb F_q$ satisfies $q\equiv 1 \pmod{\ell^2}$., Comment: 12 pages. Comments are welcome!

The linear solution for quadrupolar perturbations around de Sitter spacetime was recently constructed. In this paper, we provide the flux-balance laws for each background symmetry (dilatations, rotations, spatial translations and cosmological boosts) in terms of source moments at quadrupolar order. We write the dilatation flux-balance law in two distinct ways, which allows to contrast two distinct proposals for the negative definite energy flux. The standard Poincar\'e flux balance laws at future null infinity are recovered in the flat limit of the $SO(1,4)$ flux-balance laws., Comment: 27 pages

Automated lumbar spine segmentation is very crucial for modern diagnosis systems. In this study, we introduce a novel machine-agnostic approach for segmenting lumbar vertebrae and intervertebral discs from MRI images, employing a cascaded model that synergizes an ROI detection and a Self-organized Operational Neural Network (Self-ONN)-based encoder-decoder network for segmentation. Addressing the challenge of diverse MRI modalities, our methodology capitalizes on a unique dataset comprising images from 12 scanners and 34 subjects, enhanced through strategic preprocessing and data augmentation techniques. The YOLOv8 medium model excels in ROI extraction, achieving an excellent performance of 0.916 mAP score. Significantly, our Self-ONN-based model, combined with a DenseNet121 encoder, demonstrates excellent performance in lumbar vertebrae and IVD segmentation with a mean Intersection over Union (IoU) of 83.66%, a sensitivity of 91.44%, and Dice Similarity Coefficient (DSC) of 91.03%, as validated through rigorous 10-fold cross-validation. This study not only showcases an effective approach to MRI segmentation in spine-related disorders but also sets the stage for future advancements in automated diagnostic tools, emphasizing the need for further dataset expansion and model refinement for broader clinical applicability., Comment: 19 Pages, 11 Figures, Expert Systems with Applications, 2024

In this chapter, we aim to explore an in-depth exploration of the specialized hardware accelerators designed to enhance Artificial Intelligence (AI) applications, focusing on their necessity, development, and impact on the field of AI. It covers the transition from traditional computing systems to advanced AI-specific hardware, addressing the growing demands of AI algorithms and the inefficiencies of conventional architectures. The discussion extends to various types of accelerators, including GPUs, FPGAs, and ASICs, and their roles in optimizing AI workloads. Additionally, it touches on the challenges and considerations in designing and implementing these accelerators, along with future prospects in the evolution of AI hardware. This comprehensive overview aims to equip readers with a clear understanding of the current landscape and future directions in AI hardware development, making it accessible to both experts and newcomers to the field., Comment: The book chapter is a part of the Book, "AI-Enabled Electronic Circuit and System Design" with ISBN 978-3-031-71435-1

Organizational success in todays competitive employment market depends on choosing the right staff. This work evaluates software engineer profiles using an automated staff selection method based on advanced natural language processing (NLP) techniques. A fresh dataset was generated by collecting LinkedIn profiles with important attributes like education, experience, skills, and self-introduction. Expert feedback helped transformer models including RoBERTa, DistilBERT, and a customized BERT variation, LastBERT, to be adjusted. The models were meant to forecast if a candidate's profile fit the selection criteria, therefore allowing automated ranking and assessment. With 85% accuracy and an F1 score of 0.85, RoBERTa performed the best; DistilBERT provided comparable results at less computing expense. Though light, LastBERT proved to be less effective, with 75% accuracy. The reusable models provide a scalable answer for further categorization challenges. This work presents a fresh dataset and technique as well as shows how transformer models could improve recruiting procedures. Expanding the dataset, enhancing model interpretability, and implementing the system in actual environments will be part of future activities., Comment: 6 pages, 12 figures, conference paper

Can peer recommendation engines elevate people's creative performances in self-organizing social networks? Answering this question requires resolving challenges in data collection (e.g., tracing inspiration links and psycho-social attributes of nodes) and intervention design (e.g., balancing idea stimulation and redundancy in evolving information environments). We trained a model that predicts people's ideation performances using semantic and network-structural features in an online platform. Using this model, we built SocialMuse, which maximizes people's predicted performances to generate peer recommendations for them. We found treatment networks leveraging SocialMuse outperforming AI-agnostic control networks in several creativity measures. The treatment networks were more decentralized than the control, as SocialMuse increasingly emphasized network-structural features at large network sizes. This decentralization spreads people's inspiration sources, helping inspired ideas stand out better. Our study provides actionable insights into building intelligent systems for elevating creativity.

One of the most critical components of the Internet that an attacker could exploit is the DNS (Domain Name System) protocol and infrastructure. Researchers have been constantly developing methods to detect and defend against the attacks against DNS, specifically DNS flooding attacks. However, most solutions discard packets for defensive approaches, which can cause legitimate packets to be dropped, making them highly dependable on detection strategies. In this paper, we propose MTDNS, a resilient MTD-based approach that employs Moving Target Defense techniques through Software Defined Networking (SDN) switches to redirect traffic to alternate DNS servers that are dynamically created and run under the Network Function Virtualization (NFV) framework. The proposed approach is implemented in a testbed environment by running our DNS servers as separate Virtual Network Functions, NFV Manager, SDN switches, and an SDN Controller. The experimental result shows that the MTDNS approach achieves a much higher success rate in resolving DNS queries and significantly reduces average latency even if there is a DNS flooding attack., Comment: 6 pages, Accepted for publication at IEEE CCNC 2025

Retrieval-Augmented Generation (RAG) has been shown to enhance the factual accuracy of Large Language Models (LLMs), but existing methods often suffer from limited reasoning capabilities in effectively using the retrieved evidence, particularly when using open-source LLMs. To mitigate this gap, we introduce a novel framework, Open-RAG, designed to enhance reasoning capabilities in RAG with open-source LLMs. Our framework transforms an arbitrary dense LLM into a parameter-efficient sparse mixture of experts (MoE) model capable of handling complex reasoning tasks, including both single- and multi-hop queries. Open-RAG uniquely trains the model to navigate challenging distractors that appear relevant but are misleading. As a result, Open-RAG leverages latent learning, dynamically selecting relevant experts and integrating external knowledge effectively for more accurate and contextually relevant responses. In addition, we propose a hybrid adaptive retrieval method to determine retrieval necessity and balance the trade-off between performance gain and inference speed. Experimental results show that the Llama2-7B-based Open-RAG outperforms state-of-the-art LLMs and RAG models such as ChatGPT, Self-RAG, and Command R+ in various knowledge-intensive tasks. We open-source our code and models at https://openragmoe.github.io/, Comment: Accepted to EMNLP 2024 Findings. Website: https://openragmoe.github.io/. 14 pages, 7 figures, 5 tables