arXiv subjects

Anurag Tripathi

Publications and source records attributed to Anurag Tripathi.

At least 19 recordsLinked to original sources

arXiv2026

Intercoupling of Segregation and Rheology in Spatially Developing Granular Chute Flows

We investigate the flow and segregation of binary granular mixtures with density differences down a long chute using a continuum framework that couples a particle-force-based segregation model with an inertial-number-based local rheology. The steady-state momentum and convection-diffusion-segregation equations are solved simultaneously, explicitly accounting for the two-way coupling between segregation and flow. Predicted concentration and velocity fields at different streamwise locations are validated against representative DEM simulations. The validated model is then used to examine the influence of density ratio, mixture composition, and chute inclination on segregation over a wide range of conditions. The chute length required to achieve fully developed segregation is quantified and compared with the development length for monodisperse granular flow. At low density ratios and/or low inclinations, segregation develops over much longer distances than the velocity field. In contrast, at higher inclinations and larger density contrasts, the two length scales become comparable, demonstrating that neglecting flow development can significantly underestimate segregation evolution.

cond-mat.soft↗

arXiv2026

Particle Force-Based Continuum Model for Multicomponent Size Segregating Mixtures

We investigate size difference driven segregation in dense granular flows of multicomponent mixtures down a periodic chute using continuum model and Discrete Element Method (DEM) simulations. A previously developed particle force-based segregation model for binary mixtures is systematically extended to mixtures comprising three or more particle species differing in size. The generalized model accounts for inter-species interactions by computing the net force on each component in the presence of all others, without relying on empirical percolation velocity. This segregation model is coupled with a mixture rheology model and incorporated into the species transport and momentum balance equations to develop a continuum model that predicts the spatial and temporal evolution of species concentration and velocity fields. The continuum model predictions are found to be in agreement with DEM simulation data for ternary and quaternary mixtures over a wide range of mixture compositions and chute inclinations at moderate size ratios for well-mixed and small-near-base configurations. For larger size ratios, the one dimensional model predictions capture the qualitative segregation trend while showing relatively larger quantitative differences from DEM data. For an initial configuration, having large particles near base and small particles near the free surface, a Rayleigh-Taylor like instability at early times is observed. Due to the presence of this instability, two dimensional evolution of the species concentration fields is present for initial part of the flow. Predictions of such features requires the extension of the one dimensional continuum model to two dimensions.

cond-mat.soft↗

arXiv2026

Benchmarking Pedestrian Dynamics Models for Common Scenarios: An Evaluation of Force-Based Models

Extensive research in pedestrian dynamics has primarily focused on crowded conditions and associated phenomena, such as lane formation, evacuation, etc. Several force-based models have been developed to predict the behavior in these situations. In contrast, there is a notable gap in terms of investigations of the moderate-to-low density situations. These scenarios are extremely commonplace across the world, including the highly populated nations like India. Additionally, the details of force-based models are expected to show significant effects at these densities, whereas the crowded, nearly packed, conditions may be expected to be governed largely by contact forces. In this study, we address this gap and comprehensively evaluate the performance of different force-based models in some common scenarios. Towards this, we perform controlled experiments in four situations: avoiding a stationary obstacle, position-swapping by walking toward each other, overtaking to reach a common goal, and navigating through a maze of obstacles. The performance evaluation consists of two stages and six evaluating parameters - successful trajectories, overlapping proportion, oscillation strength, path smoothness, speed deviation, and travel time. Firstly, models must meet an eligibility criterion of at least 80\% successful trajectories and secondly, the models are scored based on the cutoff values established from the experimental data. We evaluated five force-based models where the best one scored 57.14\%. Thus, our findings reveal significant shortcomings in the ability of these models to yield accurate predictions of pedestrian dynamics in these common situations.

physics.soc-ph↗

arXiv2025

Shape Characterization of Ferrous Burden Material of Blast Furnace Feed using Image Analysis

In this study, we attempt to characterize the shape of three different types of grains commonly used in the iron and steel-making industry, namely pellet, sinter and iron ore lump. We choose particles over the entire size ranges used in industrial-scale blast furnace and consider two different size ranges of pellet particles and four different size ranges for sinter and iron ore lumps. We perform image analysis to calculate size and shape-related properties of the grains. We select some of the common length scales used to measure the size of the particles and categorize them in different classes. We show that the length scales of a particular category class are well correlated with each other. We identify the independent, uncorrelated length scales for the particles from different categories. Using these uncorrelated length scales, we define different shape descriptors and obtain the distribution of these shape descriptors for each component of the blast furnace feed. Our image analysis results show that the cumulative distribution curves for these shape descriptors turn out to be nearly independent of the size range for a given type of material. Our study identifies the three key shape descriptors that are required to characterize the shape of the blast furnace feed. Two of these shape descriptors, namely the aspect ratio and the circularity, have been considered important by the researchers earlier as well. The third shape descriptor, the average contact eccentricity to the projected particle diameter ratio, usually not considered to be an important shape descriptor in previous studies, is of high relevance for Discrete Element Method simulations of granular materials.

cond-mat.soft↗

arXiv2025

A network-based approach to measure granule size distribution for discrete element modeling of granulation

Drum granulation is a size enlargement process where granular material is agitated with a liquid binder to form larger size granules. Discrete element modeling is increasingly being used to better understand and investigate the granulation process. However, unlike experiments the measurement of granule size within a DEM framework often necessitates an explicit quantitative definition of a granule and a corresponding granule identification method. In this work, we show that the existing definitions and the associated methods in literature are ineffective at identifying granules for dense flows such as during drum granulation. We propose an improved definition and granule identification method based on community-detection used in network science literature. The proposed method better identifies granules in a drum granulator as benchmarked against liquid-settling. We also vary granulation process parameters like liquid content and fill level and study their effect on the cumulative granule size distribution attained after drum granulation. We find that the existing granule-identification methods fail to reproduce the well-known effects of process parameters on the cumulative granule size distribution. The proposed method, based on community detection, reproduces the effects with better accuracy.

cond-mat.soft↗

arXiv2025

CON-QA: Privacy-Preserving QA using cloud LLMs in Contract Domain

As enterprises increasingly integrate cloud-based large language models (LLMs) such as ChatGPT and Gemini into their legal document workflows, protecting sensitive contractual information - including Personally Identifiable Information (PII) and commercially sensitive clauses - has emerged as a critical challenge. In this work, we propose CON-QA, a hybrid privacy-preserving framework designed specifically for secure question answering over enterprise contracts, effectively combining local and cloud-hosted LLMs. The CON-QA framework operates through three stages: (i) semantic query decomposition and query-aware document chunk retrieval using a locally deployed LLM analysis, (ii) anonymization of detected sensitive entities via a structured one-to-many mapping scheme, ensuring semantic coherence while preventing cross-session entity inference attacks, and (iii) anonymized response generation by a cloud-based LLM, with accurate reconstruction of the original answer locally using a session-consistent many-to-one reverse mapping. To rigorously evaluate CON-QA, we introduce CUAD-QA, a corpus of 85k question-answer pairs generated over 510 real-world CUAD contract documents, encompassing simple, complex, and summarization-style queries. Empirical evaluations, complemented by detailed human assessments, confirm that CON-QA effectively maintains both privacy and utility, preserves answer quality, maintains fidelity to legal clause semantics, and significantly mitigates privacy risks, demonstrating its practical suitability for secure, enterprise-level contract documents.

cs.AI↗

arXiv2025

HHNAS-AM: Hierarchical Hybrid Neural Architecture Search using Adaptive Mutation Policies

Neural Architecture Search (NAS) has garnered significant research interest due to its capability to discover architectures superior to manually designed ones. Learning text representation is crucial for text classification and other language-related tasks. The NAS model used in text classification does not have a Hybrid hierarchical structure, and there is no restriction on the architecture structure, due to which the search space becomes very large and mostly redundant, so the existing RL models are not able to navigate the search space effectively. Also, doing a flat architecture search leads to an unorganised search space, which is difficult to traverse. For this purpose, we propose HHNAS-AM (Hierarchical Hybrid Neural Architecture Search with Adaptive Mutation Policies), a novel approach that efficiently explores diverse architectural configurations. We introduce a few architectural templates to search on which organise the search spaces, where search spaces are designed on the basis of domain-specific cues. Our method employs mutation strategies that dynamically adapt based on performance feedback from previous iterations using Q-learning, enabling a more effective and accelerated traversal of the search space. The proposed model is fully probabilistic, enabling effective exploration of the search space. We evaluate our approach on the database id (db_id) prediction task, where it consistently discovers high-performing architectures across multiple experiments. On the Spider dataset, our method achieves an 8% improvement in test accuracy over existing baselines.

cs.LG↗

arXiv2025

End-to-End Text-to-SQL with Dataset Selection: Leveraging LLMs for Adaptive Query Generation

Text-to-SQL bridges the gap between natural language and structured database language, thus allowing non-technical users to easily query databases. Traditional approaches model text-to-SQL as a direct translation task, where a given Natural Language Query (NLQ) is mapped to an SQL command. Recent advances in large language models (LLMs) have significantly improved translation accuracy, however, these methods all require that the target database is pre-specified. This becomes problematic in scenarios with multiple extensive databases, where identifying the correct database becomes a crucial yet overlooked step. In this paper, we propose a three-stage end-to-end text-to-SQL framework to identify the user's intended database before generating SQL queries. Our approach leverages LLMs and prompt engineering to extract implicit information from natural language queries (NLQs) in the form of a ruleset. We then train a large db\_id prediction model, which includes a RoBERTa-based finetuned encoder, to predict the correct Database identifier (db\_id) based on both the NLQ and the LLM-generated rules. Finally, we refine the generated SQL by using critic agents to correct errors. Experimental results demonstrate that our framework outperforms the current state-of-the-art models in both database intent prediction and SQL generation accuracy.

cs.LG↗

arXiv2025

CWebGen -- A tool to study colour structure of scattering amplitudes in IR limit

Infrared singularities in perturbative Quantum Chromodynamics (QCD) are captured by the Soft function, which can be calculated efficiently using Feynman diagrams known as webs. The starting point for calculating Soft function using webs is to compute the web mixing matrices using a well known replica trick algorithm. We present a package implemented in Mathematica to calculate these mixing matrices. Along with the package, we provide several state-of-the art computations.

hep-ph↗

arXiv2025

Transient segregation of bi-disperse granular mixtures in a periodic chute flow

Transient size segregation of a bi-disperse granular mixture flowing over a periodic chute is studied using the Discrete Element Method and continuum simulations. A recently developed particle force-based size segregation model is used to predict the time-dependent flow properties of binary mixtures starting from rest. A two-way coupled continuum model that solves the momentum balance and convection-diffusion equations by incorporating the mixture segregation model along with the generalized inertial number-based rheological model is developed for predicting the evolution of segregation. The predicted concentration profiles and other flow properties of the mixture are found to be in good agreement with the DEM data for a variety of compositions. The evolution of the centre of mass of the two species with time is also very well captured for different initial configurations and size ratios using the particle force-based segregation model.

cond-mat.soft↗

arXiv2025

Enhanced UV Photodetector Efficiency with a ZnO/Ga$_2$O$_3$ Heterojunction

Heterostructures comprising uncoated ZnO and coated with thin layers of Ga$_2$O$_3$ were produced using spin-coating and subsequent hydrothermal processing. X-ray diffraction examination verifies the structural integrity of the synthesized heterostructures (HTs). Optical and photoluminescence spectra were recorded to assess the variation in absorption and emission of the Ga$_2$O$_3$-coated HTs in comparison to the pristine ZnO. We conducted comparative density-functional theory (DFT) computations to corroborate the measured band gaps of both categories of HTs. To assess the stability of our devices, the transient response to on/off light switching under zero bias has been studied. The rise time $τ_{r1}$ ($τ_{r2}$) is 2300 (500) ms and the decay time $τ_{d1}$ ($τ_{d2}$) is 2700 (5000) ms have been observed for bare ZnO and ZnO/Ga$_2$O$_3$ HTs, respectively. A significant amount of change was also observed in the electrical transport properties from bare ZnO to ZnO/Ga$_2$O$_3$. To see the performance of device, responsivity (R) and detectivity (D = 1/NEP$_B$) have been measured. It is evident from observation that responsivity of a device shows maximum value in UV region while it is reducing with visible region for HTs. In case of detectivity, the maximum value reached was $145 \times 10^{14}$ Hz$^{1/2}$/W (at ~ 200 nm) and $38 \times 10^{14}$ Hz$^{1/2}$/W (at 300 nm) for Ga$_2$O$_3$ coated ZnO, and bare ZnO HTs, respectively. The maximum responsivity measured for the bare ZnO HTs is 7 (A/W) while that of Ga$_2$O$_3$ coated ZnO HTs is 38 (A/W). It suggests a simple way of designing materials for fabricating broad-range cost-effective photodetectors.

cond-mat.mtrl-sci↗

arXiv2025

Colour structure of next-to-eikonal correlator webs at three loops

Correlators of Wilson-line, which capture eikonal contributions, are known to exponentiate in non-abelian gauge theories, and their logarithms can be organised in terms of collections of Feynman diagrams called webs. In~\cite{Agarwal:2020nyc} the concept of correlator web (Cweb), which is a set of skeleton diagrams built with connected gluon correlators and provides a generalisation of webs was introduced. The part of next-to-eikonal contributions to the scattering amplitude which exponentiates is given in terms of next-to-eikonal Cwebs~\cite{Gardi:2010rn,Laenen:2008gt}. In the present article we study next-to-eikonal Cwebs at three loop order, the order at which non trivial Cweb mixing matrices appear for the first time. The methods developed in~\cite{Agarwal:2022wyk} to construct mixing matrices directly without use of replica trick are used in this article to obtain the mixing matrices for next-to-eikonal Cwebs after we establish a relationship between next-to-eikonal and eikonal Cwebs.

hep-ph↗

arXiv2025

Information Extraction from Visually Rich Documents using LLM-based Organization of Documents into Independent Textual Segments

Information extraction (IE) from Visually Rich Documents (VRDs) containing layout features along with text is a critical and well-studied task. Specialized non-LLM NLP-based solutions typically involve training models using both textual and geometric information to label sequences/tokens as named entities or answers to specific questions. However, these approaches lack reasoning, are not able to infer values not explicitly present in documents, and do not generalize well to new formats. Generative LLM-based approaches proposed recently are capable of reasoning, but struggle to comprehend clues from document layout especially in previously unseen document formats, and do not show competitive performance in heterogeneous VRD benchmark datasets. In this paper, we propose BLOCKIE, a novel LLM-based approach that organizes VRDs into localized, reusable semantic textual segments called $\textit{semantic blocks}$, which are processed independently. Through focused and more generalizable reasoning,our approach outperforms the state-of-the-art on public VRD benchmarks by 1-3% in F1 scores, is resilient to document formats previously not encountered and shows abilities to correctly extract information not explicitly present in documents.

cs.IR↗

arXiv2025

Transient size segregation of binary granular mixtures

Transient size segregation of a bi-disperse granular mixture flowing over a periodic chute is studied using DEM simulations and theory. A recently developed particle force-based size segregation model has been shown to successfully predict the steady state behavior of binary granular mixtures [1]. This promising model is used to predict the time-dependent segregation of different size binary mixtures in this work. A one dimensional continuum model is developed to solve the convection-diffusion equation by incorporating a mixture segregation model along with rheological model. The inter-coupling of segregation with rheology is accounted to predict evolution of species concentration. We also investigate the effect of different initial configurations (Large near base (LNB), Small near base (SNB) and well-mixed) on the transient evolution of the flow and segregation. The particle force-based segregation model is able to predict the evolution of the concentration profile for all three initial configurations for smallest size ratio of 1.25. Significant deviations, however, are observable for larger size ratios, suggesting the need to account for the evolution of the velocity field in the model.

cond-mat.soft↗

arXiv2025

Variable Goal Approach (VGA) Enhancing Pedestrian Dynamics Modeling

Pedestrian dynamics models have provided valuable insights into pedestrian interactions, collision avoidance, and self-organized crowd behavior using mathematical, computational, AI-based, and heuristic approaches. However, existing models often fail to capture fundamental aspects of human decision-making, particularly the tendency to adopt indirect routes by sequentially selecting intermediate goals within the line of sight. In this study, we propose a novel Variable Goal Approach (VGA) that integrates human intelligence into pedestrian dynamics models by introducing multiple intermediate goals, termed variable goals, which guide pedestrians toward their final destination. These variable goals function as an adaptive guidance mechanism, enabling smoother transitions and dynamic navigation. VGA also enhances the efficiency of a model while minimizing interactions and disruptions. By strategically positioning variable goals, VGA introduces an element of stochasticity. This allows the model to simulate varied pedestrian paths under identical conditions, reflecting the diversity in human decision-making. In addition to its effectiveness in simple scenarios, VGA demonstrates strong performance in replicating high-density scenarios, such as lane formation, providing results that closely match real-world data.

physics.soc-ph↗

arXiv2025

Automatic Estimation of Pedestrian Gait Features using a single camera recording: Algorithm and Statistical Analysis for Gender Difference and Obstacle Interactions

The pedestrian gait features - body sway frequency, amplitude, stride length, and speed, along with pedestrian personal space and directional bias, are important parameters to be used in different pedestrian dynamics studies. Gait feature measurements are paramount for wide-ranging applications, varying from the medical field to the design of bridges. Personal space and choice of direction (directional bias) play important roles during crowd simulations. In this study, we formulate an automatic algorithm for calculating the gait features of a trajectory extracted from video recorded using a single camera attached to the roof of a building. Our findings indicate that females have 28.64% smaller sway amplitudes, 8.68% smaller stride lengths, and 8.14% slower speeds compared to males, with no significant difference in frequency. However, according to further investigation, our study reveals that the body parameters are the main variables that dominate gait features rather than gender. We have conducted three experiments in which the volunteers are walking towards the destination a) without any obstruction, b) with a stationary non-living obstacle present in the middle of the path, and c) with a human being standing in the middle of the path. From a comprehensive statistical analysis, key observations include no significant difference in gait features with respect to gender, no significant difference in gait features in the absence or presence of an obstacle, pedestrians treating stationary human beings and stationary obstacles the same given that the gender is same to match the comfort level, and a directional bias towards the left direction, likely influenced by left-hand traffic rule in India.

physics.soc-ph↗

arXiv2024

Information Extraction from Heterogeneous Documents without Ground Truth Labels using Synthetic Label Generation and Knowledge Distillation

Invoices and receipts submitted by employees are visually rich documents (VRDs) with textual, visual and layout information. To protect against the risk of fraud and abuse, it is crucial for organizations to efficiently extract desired information from submitted receipts. This helps in the assessment of key factors such as appropriateness of the expense claim, adherence to spending and transaction policies, the validity of the receipt, as well as downstream anomaly detection at various levels. These documents are heterogeneous, with multiple formats and languages, uploaded with different image qualities, and often do not contain ground truth labels for the efficient training of models. In this paper we propose Task Aware Instruction-based Labelling (TAIL), a method for synthetic label generation in VRD corpuses without labels, and fine-tune a multimodal Visually Rich Document Understanding Model (VRDU) on TAIL labels using response-based knowledge distillation without using the teacher model's weights or training dataset to conditionally generate annotations in the appropriate format. Using a benchmark external dataset where ground truth labels are available, we demonstrate conditions under which our approach performs at par with Claude 3 Sonnet through empirical studies. We then show that the resulting model performs at par or better on the internal expense documents of a large multinational organization than state-of-the-art LMM (large multimodal model) Claude 3 Sonnet while being 85% less costly and ~5X faster, and outperforms layout-aware baselines by more than 10% in Average Normalized Levenshtein Similarity (ANLS) scores due to its ability to reason and extract information from rare formats. Finally, we illustrate the usage of our approach in overpayment prevention.

cs.CL↗

arXiv2024

Transient segregation of different density granular mixtures

We study time-dependent density segregation of granular mixtures flowing over an inclined plane. Discrete Element Method (DEM) simulations in a periodic box are performed for granular mixtures of same size and different density particles flowing under the influence of gravity. In addition, a continuum model is developed to solve the momentum balance equations along with species transport equation by accounting for the inter-coupling of segregation and rheology. The particle force-based density segregation theory has been used along with the $μ-I$ rheology to predict evolution of flow properties with time for binary and multicomponent mixtures. The effect of particle arrangements on the transient evolution of flow properties for three different initial configurations is investigated using both continuum and DEM simulations. Continuum predictions for various flow properties of interest such as species concentration, velocity, pressure, and shear stress at different time instants are compared with DEM simulations. The results from the discrete and continuum models are found to be in good agreement with each other for well-mixed and heavy-near-base initial configurations. However, the continuum model is unable to predict the flow evolution for the light-near-base initial configuration. DEM simulations reveal the presence of an instability driven, quick segregation for this configuration which is not predicted by the one dimensional model and requires generalization to three dimensions.

cond-mat.soft↗