arXiv subjects

Shubham Sharma

Publications and source records attributed to Shubham Sharma.

At least 19 recordsLinked to original sources

arXiv2026

Extraction of Pion Unpolarized Quark and Gluon Generalized Parton Distributions using Deep Neural-Networks

We present a deep neural-network (DNN) extraction of the pion unpolarized quark and gluon generalized parton distributions (GPDs) using the corresponding parton distribution functions (PDFs) from the JAM21 and xFitter analysis, together with experimental measurements of the pion electromagnetic form factor (EMFF) and lattice quantum chromodynamics (QCD) results. The GPDs are parameterized using a physics-informed neural-network (PINN) that incorporates the known PDF behavior, an exponential momentum-transfer dependence, and a trainable neural network (NN) component. The network parameters are determined by minimizing a $\chi^2$-based loss function. For the valence-quark GPDs, the loss function includes contributions from the EMFF, squared EMFF, charge-normalization constraints, and regularization terms. For the gluon GPDs, it incorporates constraints from the gluon gravitational form factors together with regularization. This framework enables a flexible, nonparametric extraction while preserving the essential theoretical and phenomenological constraints. By employing the full ensemble of available PDF replicas, we quantify the uncertainties of the extracted GPDs over a broad kinematic range in the longitudinal momentum fraction and momentum transfer, with the uncertainty bands corresponding to the $1\sigma$ confidence interval. The extracted valence-quark GPDs are found to be in good agreement with available lattice-QCD calculations. Our study demonstrates that DNN-based methods provide a flexible and robust framework for extracting pion GPDs and probing the multidimensional internal structure of the pion, offering a promising avenue for future investigations of hadron tomography.

hep-ph↗

arXiv2026

Device Invariance using Domain Adaptation on Acoustic Scene Classification

This paper explores the effectiveness of domain adaptation techniques when using convolutional neural network (CNN)-based and transformer-based feature representations for acoustic scene classification. Two well-known domain adaptation techniques, namely domain adversarial neural network (also called DANN) and conditional domain adversarial network (also called CDAN) are evaluated under various domain shifts. Our study indicates that DANN provides effective domain adaptation fairly consistently for both feature extractors. On the other hand, CDAN provides effective domain adaptation only for CNN-based feature extractors. The study gives insights into how domain adaptation methods may need to be tailored to the underlying feature representation. Experimental evaluation with multiple devices on the DCASE 2020 dataset supports the observations.

eess.AS↗

arXiv2026

Dyna-Mat: End-to-end benchmarking of foundation machine learning interatomic potentials in finite-temperature ensembles

Foundation machine learning interatomic potentials (MLIPs) are increasingly being used as drop-in replacements for first-principles calculations, enabling simulations of materials at length and time scales that were previously inaccessible. However, due to lack of ground truth data, their accuracy on structural and dynamical observables in finite thermodynamic ensembles is yet to be established. Here, we introduce Dyna-Mat-v1.0, a benchmark dataset of condensed-phase first-principles molecular dynamics trajectories designed to test foundation MLIPs at realistic finite-temperature conditions. Using this dataset, we evaluate 15 foundation MLIPs across four model tiers by comparing both single-point energy and force errors on first-principles configurations and observables generated from MLIP-driven trajectories. We find that "on average" models with lower single-point force errors also yield lower errors for structural and dynamical observables. However, there are individual systems for which low force errors lead to qualitative failures in the predicted structure. Pressure remains poorly described across most models, pointing to limitations in the density functional theory stress labels available in current large-scale training datasets. Finally, we construct an accuracy-cost Pareto frontier to identify the best trade-offs for molecular dynamics with foundation MLIPs, finding that the latest generation of cross-trained models is close to Pareto-optimal according to the accuracy metrics considered here. Overall, Dyna-Mat-v1.0 shows that end-to-end finite-temperature validation is essential for quantifying the predictive behaviour of foundation MLIPs, and provides a simple, scalable route for assessing them beyond static and harmonic benchmarks relevant to materials design.

cond-mat.mtrl-sci↗

arXiv2026

PulseCX: Breaking the Closed-World Assumption in Real-Time CX

Conversational AI agents in Customer Experience (CX) typically suffer from a Closed-World Constraint, ignoring high-velocity external shifts like viral trends or outages. Ad-hoc web search attempts to bridge this gap but often introduce prohibitive latency and context poisoning. We introduce PulseCX, a framework that decouples knowledge acquisition from consumption. Adopting a structure-first paradigm, PulseCX employs an asynchronous agent to linearize signals into a Decay-Aware Temporal Knowledge Graph (DA-TKG) governed by reinforcement--decay dynamics to actively manage information lifecycles. By coupling this self-evolving memory with hierarchical intent gating, PulseCX removes synchronous search bottlenecks (<10ms overhead) and drives significant gains in Intent Resolution (IRR) and Customer Satisfaction (s-CSAT) in dynamic environments.

cs.AI↗

arXiv2026

Role of higher twist distributions in the tomography of proton

We have studied the higher-twist distributions of the proton, including T-even and T-odd transverse momentum-dependent parton distributions (TMDs). Under the umbrella of the light-front framework, we have chosen two distinctive approaches of quark-spectator systems for comparison, one inspired by the soft-wall AdS/QCD and another with a dipolar form factor at the nucleon-quark-diquark vertex. The comprehensive picture at higher-twist provided by both T-even and T-odd TMDs not only aids deeper insights into the internal structure of the proton in the quark sector but also provides an interpretation of different components of the energy-momentum tensor in quantum chromodynamics. Hence, using these standard parton distribution functions, further predictions regarding the physical insights of gravitational TMDs in momentum space are also provided.

hep-ph↗

arXiv2026

Moderate-to-Large-$x$ Gluon Helicity from $J/\psi$ Production at $\sqrt{s}=27~\mathrm{GeV}$

We present a feasibility study of the longitudinal double-spin asymmetry $A_{LL}$ in inclusive $J/\psi$ production in polarized proton-proton collisions at $\sqrt{s}\approx 27~\mathrm{GeV}$ at the Spin Physics Detector (SPD) of the Nuclotron-based Ion Collider fAcility (NICA). At these moderate energies, $J/\psi$ production is dominated by gluon-gluon fusion, probing gluon momentum fractions $x\approx 0.1$-$0.2$ at central rapidity and highly asymmetric configurations at forward rapidity, where one parton can reach $x\approx 0.5$-$0.9$. This provides direct sensitivity to the poorly constrained moderate- to large-$x$ region of the gluon helicity distribution $\Delta g(x)$. We estimate $A_{LL}$ as a function of transverse momentum and rapidity using polarized parton distribution functions, focusing on the underlying partonic spin asymmetry. Nonperturbative long-distance effects are treated in a simplified manner and largely cancel in the asymmetry, enabling a direct assessment of gluon polarization sensitivity. We find asymmetries reaching $|A_{LL}|\approx 0.09$ at $p_T=3~\mathrm{GeV}$, with enhanced sensitivity at forward rapidity. The dominant theoretical uncertainty arises from polarized parton distribution functions. These results demonstrate that inclusive $J/\psi$ measurements at SPD/NICA provide a sensitive and complementary probe of gluon polarization at moderate and large $x$, extending constraints from RHIC into a kinematic regime not directly accessible to the EIC.

hep-ph↗

arXiv2026

Mechanical properties of proton in the momentum space

We study the parametrization of the energy-momentum tensor for the case of a proton in momentum space in terms of gravitational transverse momentum-dependent distributions (TMDs). These gravitational TMDs are investigated with the inclusion of higher-twist contributions to predict the mechanical properties, specifically the transverse pressure and shear force distributions, along with the polarization-dependent $\Pi^q_S$ and $\Pi^q_A$ terms. The corresponding distributions are computed individually for both $u$ and $d$ quark flavors. The calculations have been performed in the light-cone framework using the spectator diquark model. A strong binding contribution to the transverse pressure is observed in the low-momentum space for both quark flavors of the proton.

hep-ph↗

arXiv2026

Extraction of Pion Unpolarized Quark Generalized Parton Distribution from Charge Form Factors

Based on a global fit to experimental measurements of the pion electromagnetic form factor and parton distribution functions (PDFs), we report a data-driven determination of the unpolarized quark generalized parton distributions (GPDs) for the case of pion in the zero-skewness limit ($\xi = 0$). The form factor is parameterized using a flexible functional form constrained by data and embedded into a GPD framework constructed from collinear PDFs and a profile function encoding transverse dynamics. This approach provides a unified description of the pion's electromagnetic structure and its spatial parton distributions. We present the extracted pion GPDs and their impact-parameter-space interpretations, offering new insights into the internal structure of the lightest QCD bound state and providing essential input for future electron-ion collider studies via the Sullivan process, as well as for the exclusive $\pi^+$ electroproduction at the 12~GeV Jefferson Lab program, pion-induced exclusive measurements at COMPASS, proposed pion-beam experiments at AMBER, and phenomenological and lattice investigations of the structure of the meson.

hep-ph↗

arXiv2026

The Eleventh NTIRE 2026 Efficient Super-Resolution Challenge Report

This paper reviews the NTIRE 2026 challenge on efficient single-image super-resolution with a focus on the proposed solutions and results. The aim of this challenge is to devise a network that reduces one or several aspects, such as runtime, parameters, and FLOPs, while maintaining PSNR of around 26.90 dB on the DIV2K_LSDIR_valid dataset, and 26.99 dB on the DIV2K_LSDIR_test dataset. The challenge had 95 registered participants, and 15 teams made valid submissions. They gauge the state-of-the-art results for efficient single-image super-resolution.

cs.CV↗

arXiv2025

Quantum Machine Learning for Cybersecurity: A Taxonomy and Future Directions

The increasing number of cyber threats and rapidly evolving tactics, as well as the high volume of data in recent years, have caused classical machine learning, rules, and signature-based defence strategies to fail, rendering them unable to keep up. An alternative, Quantum Machine Learning (QML), has recently emerged, making use of computations based on quantum mechanics. It offers better encoding and processing of high-dimensional structures for certain problems. This survey provides a comprehensive overview of QML techniques relevant to the domain of security, such as Quantum Neural Networks (QNNs), Quantum Support Vector Machines (QSVMs), Variational Quantum Circuits (VQCs), and Quantum Generative Adversarial Networks (QGANs), and discusses the contributions of this paper in relation to existing research in the field and how it improves over them. It also maps these methods across supervised, unsupervised, and generative learning paradigms, and to core cybersecurity tasks, including intrusion and anomaly detection, malware and botnet classification, and encrypted-traffic analytics. It also discusses their application in the domain of cloud computing security, where QML can enhance secure and scalable operations. Many limitations of QML in the domain of cybersecurity have also been discussed, along with the directions for addressing them.

cs.LG↗

arXiv2025

Cache What Lasts: Token Retention for Memory-Bounded KV Cache in LLMs

Memory and computation remain core bottlenecks in long-horizon LLM inference due to the quadratic cost of self-attention and the ever-growing key-value (KV) cache. Existing strategies for memory-bounded inference, such as quantization, offloading, or heuristic KV eviction, either incur high orchestration costs or rely on unreliable attention-based proxies of importance. We propose TRIM-KV, a novel approach that learns each token's intrinsic importance at creation time via a lightweight retention gate. Each gate predicts a scalar retention score that decays over time, reflecting the long-term utility of the token for a specific layer and head. Tokens with low scores are evicted when the memory budget is exceeded, ensuring that the cache always contains the most critical tokens. TRIM-KV is trained efficiently through distillation from a frozen LLM combined with a capacity loss, requiring only gate fine-tuning and adding negligible inference overhead. Across mathematical reasoning (GSM8K, MATH-500, AIME24), procedural generation (LongProc), conversational long-memory benchmarks (LongMemEval), and long-context understanding (LongBenchV2 and SCBench), TRIM-KV consistently outperforms strong eviction and learnable retrieval baselines, especially in low-memory regimes. Remarkably, it even surpasses full-cache models in some settings, showing that selective retention can serve as a form of regularization, suppressing noise from uninformative tokens. Qualitative analyses further reveal that learned retention scores align with human intuition, naturally recovering heuristics such as sink tokens, sliding windows, and gist compression without explicit design. Beyond efficiency, retention scores provide insights into layer- and head-specific roles, suggesting a new path toward LLM interpretability.

cs.LG↗

arXiv2025

Valence quark distribution of rho meson using light-front quark model

We investigate the partonic structure of the $\rho$ meson, the lightest spin-$1$ vector meson, within the light-front quark model (LFQM). To explore the sensitivity to model assumptions, we employ two distinct types of spin wave functions in the LFQM. Using light-front helicity wave functions, we derive explicit expressions for the leading-twist and subleading-twist quark parton distribution functions (PDFs), and evolve the leading-twist PDFs to higher scales with next-to-leading order (NLO) Dokshitzer--Gribov--Lipatov--Altarelli--Parisi (DGLAP) evolution. We have also calculated the Mellin moment from the evolved PDFs using a simple neural network frame and compared with available theoretical predictions. Furthermore, we compute the full set of nine leading-twist transverse-momentum-dependent distributions (TMDs) for the valence quark in the $\rho$ meson, including three tensor TMDs that arise from spin-$1$ tensor polarization of the hadron. Positivity constraints for the PDFs and TMDs are examined within this framework. Our findings highlight the crucial role of tensor polarization in shaping the three-dimensional partonic structure of vector mesons.

hep-ph↗

arXiv2025

$J/\psi$ production in proton-proton collisions at Spin Physics Detector energies of the JINR Nuclotron-based Ion Collider fAcility

We investigate inclusive $J/\psi$ production in proton-proton collisions at tens of GeV $\sqrt{s}$ energy, relevant for forthcoming measurements with the Spin Physics Detector (SPD) at NICA. Simulations are performed using the PEGASUS event generator with transverse-momentum-dependent (TMD) gluon densities, comparing the recent KMR-based KL$'2025$ and CCFM-based LLM$'2024$ parametrizations. Differential cross sections in rapidity and transverse momentum exhibit smooth, stable behavior under renormalization-scale variation. Normalized $p_T$ spectra reveal distinct hardening patterns linked to the underlying gluon $k_T$ broadening in each model. The relative contributions of color-singlet and color-octet channels are also quantified, demonstrating the dominance of color-octet mechanisms in the SPD energy regime. These results provide the first detailed assessment of quarkonium production sensitivity to gluon TMDs near threshold, offering timely theoretical guidance for upcoming $J/\psi$ measurements at SPD/NICA.

hep-ph↗

arXiv2025

EchoLSTM: A Self-Reflective Recurrent Network for Stabilizing Long-Range Memory

Standard Recurrent Neural Networks, including LSTMs, struggle to model long-range dependencies, particularly in sequences containing noisy or misleading information. We propose a new architectural principle, Output-Conditioned Gating, which enables a model to perform self-reflection by modulating its internal memory gates based on its own past inferences. This creates a stabilizing feedback loop that enhances memory retention. Our final model, the EchoLSTM, integrates this principle with an attention mechanism. We evaluate the EchoLSTM on a series of challenging benchmarks. On a custom-designed Distractor Signal Task, the EchoLSTM achieves 69.0% accuracy, decisively outperforming a standard LSTM baseline by 33 percentage points. Furthermore, on the standard ListOps benchmark, the EchoLSTM achieves performance competitive with a modern Transformer model, 69.8% vs. 71.8%, while being over 5 times more parameter-efficient. A final Trigger Sensitivity Test provides qualitative evidence that our model's self-reflective mechanism leads to a fundamentally more robust memory system.

cs.LG↗

arXiv2025

Analyticup E-commerce Product Search Competition Technical Report from Team Tredence_AICOE

This study presents the multilingual e-commerce search system developed by the Tredence_AICOE team. The competition features two multilingual relevance tasks: Query-Category (QC) Relevance, which evaluates how well a user's search query aligns with a product category, and Query-Item (QI) Relevance, which measures the match between a multilingual search query and an individual product listing. To ensure full language coverage, we performed data augmentation by translating existing datasets into languages missing from the development set, enabling training across all target languages. We fine-tuned Gemma-3 12B and Qwen-2.5 14B model for both tasks using multiple strategies. The Gemma-3 12B (4-bit) model achieved the best QC performance using original and translated data, and the best QI performance using original, translated, and minority class data creation. These approaches secured 4th place on the final leaderboard, with an average F1-score of 0.8857 on the private test set.

cs.IR↗

arXiv2025

Investigating Anharmonicities in Polarization-Orientation Raman Spectra of Acene Crystals with Machine Learning

We present a first-principles machine-learning computational framework to investigate anharmonic effects in polarization-orientation (PO) Raman spectra of molecular crystals, focusing on anthracene and naphthalene. By combining machine learning models for interatomic potentials and polarizability tensors, we enable efficient, large-scale simulations that capture temperature-dependent vibrational dynamics beyond the harmonic approximation. Our approach reproduces key qualitative features observed experimentally. We show, systematically, what are the fingerprints of anharmonic lattice dynamics, thermal expansion, and Raman tensor symmetries on PO-Raman intensities. However, we find that the simulated polarization dependence of Raman intensities shows only subtle deviations from quasi-harmonic predictions, failing to capture the pronounced temperature-dependent changes that have been reported experimentally in anthracene. We propose that part of these inconsistencies stem from the impossibility to deconvolute certain vibrational peaks when only experimental data is available. This work therefore provides a foundation to improve the interpretation of PO-Raman experiments in complex molecular crystals with the aid of theoretical simulations.

cond-mat.mtrl-sci↗

arXiv2025

EduVidQA: Generating and Evaluating Long-form Answers to Student Questions based on Lecture Videos

As digital platforms redefine educational paradigms, ensuring interactivity remains vital for effective learning. This paper explores using Multimodal Large Language Models (MLLMs) to automatically respond to student questions from online lectures - a novel question answering task of real world significance. We introduce the EduVidQA Dataset with 5252 question-answer pairs (both synthetic and real-world) from 296 computer science videos covering diverse topics and difficulty levels. To understand the needs of the dataset and task evaluation, we empirically study the qualitative preferences of students, which we provide as an important contribution to this line of work. Our benchmarking experiments consist of 6 state-of-the-art MLLMs, through which we study the effectiveness of our synthetic data for finetuning, as well as showing the challenging nature of the task. We evaluate the models using both text-based and qualitative metrics, thus showing a nuanced perspective of the models' performance, which is paramount to future work. This work not only sets a benchmark for this important problem, but also opens exciting avenues for future research in the field of Natural Language Processing for Education.

cs.CL↗

arXiv2025

Geometric Mixture Classifier (GMC): A Discriminative Per-Class Mixture of Hyperplanes

Many real world categories are multimodal, with single classes occupying disjoint regions in feature space. Classical linear models (logistic regression, linear SVM) use a single global hyperplane and perform poorly on such data, while high-capacity methods (kernel SVMs, deep nets) fit multimodal structure but at the expense of interpretability, heavier tuning, and higher computational cost. We propose the Geometric Mixture Classifier (GMC), a discriminative model that represents each class as a mixture of hyperplanes. Within each class, GMC combines plane scores via a temperature-controlled soft-OR (log-sum-exp), smoothly approximating the max; across classes, standard softmax yields probabilistic posteriors. GMC optionally uses Random Fourier Features (RFF) for nonlinear mappings while keeping inference linear in the number of planes and features. Our practical training recipe: geometry-aware k-means initialization, silhouette-based plane budgeting, alpha annealing, usage-aware L2 regularization, label smoothing, and early stopping, makes GMC plug-and-play. Across synthetic multimodal datasets (moons, circles, blobs, spirals) and tabular/image benchmarks (iris, wine, WDBC, digits), GMC consistently outperforms linear baselines and k-NN, is competitive with RBF-SVM, Random Forests, and small MLPs, and provides geometric introspection via per-plane and class responsibility visualizations. Inference scales linearly in planes and features, making GMC CPU-friendly, with single-digit microsecond latency per example, often faster than RBF-SVM and compact MLPs. Post-hoc temperature scaling reduces ECE from about 0.06 to 0.02. GMC thus strikes a favorable balance of accuracy, interpretability, and efficiency: it is more expressive than linear models and lighter, more transparent, and faster than kernel or deep models.

cs.LG↗