arXiv subjects

Amit Agarwal

Publications and source records attributed to Amit Agarwal.

At least 55 records · Page 3Linked to original sources

arXiv2025

FS-DAG: Few Shot Domain Adapting Graph Networks for Visually Rich Document Understanding

In this work, we propose Few Shot Domain Adapting Graph (FS-DAG), a scalable and efficient model architecture for visually rich document understanding (VRDU) in few-shot settings. FS-DAG leverages domain-specific and language/vision specific backbones within a modular framework to adapt to diverse document types with minimal data. The model is robust to practical challenges such as handling OCR errors, misspellings, and domain shifts, which are critical in real-world deployments. FS-DAG is highly performant with less than 90M parameters, making it well-suited for complex real-world applications for Information Extraction (IE) tasks where computational resources are limited. We demonstrate FS-DAG's capability through extensive experiments for information extraction task, showing significant improvements in convergence speed and performance compared to state-of-the-art methods. Additionally, this work highlights the ongoing progress in developing smaller, more efficient models that do not compromise on performance. Code : https://github.com/oracle-samples/fs-dag

cs.CV↗

arXiv2025

Can LLMs Narrate Tabular Data? An Evaluation Framework for Natural Language Representations of Text-to-SQL System Outputs

In modern industry systems like multi-turn chat agents, Text-to-SQL technology bridges natural language (NL) questions and database (DB) querying. The conversion of tabular DB results into NL representations (NLRs) enables the chat-based interaction. Currently, NLR generation is typically handled by large language models (LLMs), but information loss or errors in presenting tabular results in NL remains largely unexplored. This paper introduces a novel evaluation method - Combo-Eval - for judgment of LLM-generated NLRs that combines the benefits of multiple existing methods, optimizing evaluation fidelity and achieving a significant reduction in LLM calls by 25-61%. Accompanying our method is NLR-BIRD, the first dedicated dataset for NLR benchmarking. Through human evaluations, we demonstrate the superior alignment of Combo-Eval with human judgments, applicable across scenarios with and without ground truth references.

cs.CL↗

arXiv2025

Aligning LLMs for Multilingual Consistency in Enterprise Applications

Large language models (LLMs) remain unreliable for global enterprise applications due to substantial performance gaps between high-resource and mid/low-resource languages, driven by English-centric pretraining and internal reasoning biases. This inconsistency undermines customer experience and operational reliability in multilingual settings such as customer support, content moderation, and information retrieval. Even with advanced Retrieval-Augmented Generation (RAG) systems, we observe up to an 29% accuracy drop in non-English languages compared to English. We propose a practical, batch-wise alignment strategy for fine-tuning LLMs, leveraging semantically equivalent multilingual data in each training batch to directly align model outputs across languages. This approach improves non-English accuracy by up to 23.9% without compromising English performance, model reasoning, or retrieval quality. Our method is simple to implement, scalable, and integrates seamlessly with existing LLM training & deployment pipelines, enabling more robust and equitable multilingual AI solutions in industry.

cs.CL↗

arXiv2025

Who's Asking? Investigating Bias Through the Lens of Disability Framed Queries in LLMs

Large Language Models (LLMs) routinely infer users demographic traits from phrasing alone, which can result in biased responses, even when no explicit demographic information is provided. The role of disability cues in shaping these inferences remains largely uncharted. Thus, we present the first systematic audit of disability-conditioned demographic bias across eight state-of-the-art instruction-tuned LLMs ranging from 3B to 72B parameters. Using a balanced template corpus that pairs nine disability categories with six real-world business domains, we prompt each model to predict five demographic attributes - gender, socioeconomic status, education, cultural background, and locality - under both neutral and disability-aware conditions. Across a varied set of prompts, models deliver a definitive demographic guess in up to 97\% of cases, exposing a strong tendency to make arbitrary inferences with no clear justification. Disability context heavily shifts predicted attribute distributions, and domain context can further amplify these deviations. We observe that larger models are simultaneously more sensitive to disability cues and more prone to biased reasoning, indicating that scale alone does not mitigate stereotype amplification. Our findings reveal persistent intersections between ableism and other demographic stereotypes, pinpointing critical blind spots in current alignment strategies. We release our evaluation framework and results to encourage disability-inclusive benchmarking and recommend integrating abstention calibration and counterfactual fine-tuning to curb unwarranted demographic inference. Code and data will be released on acceptance.

cs.CL↗

arXiv2025

Deterministic Switching in Altermagnets via Asymmetric Sublattice Spin Current

We demonstrate a deterministic switching mechanism in collinear altermagnets driven by asymmetric sublattice spin currents. Unlike conventional antiferromagnets, where combined parity-time-reversal symmetry enforces purely staggered sublattice spin torques, altermagnets host symmetry-protected nonrelativistic spin splitting that produces unequal torques on the two sublattices. Using doped FeSb$_2$ as a representative $d$-wave altermagnet, our Landau--Lifshitz--Gilbert simulations show that these torques enable magnetic-field-free and deterministic 180$^\circ$ Néel vector reversal over picosecond timescale. The mechanism is generic to even-parity altermagnets and remains effective even in centrosymmetric, weak spin-orbit coupled systems, where the Néel spin-orbit torque mechanism fails. Our results establish an experimentally accessible mechanism for switching of altermagnetic order, opening pathways for realizing ultrafast, low-power altermagnet spintronic devices.

cond-mat.mes-hall↗

arXiv2025

Atomistic origin of low thermal conductivity in quaternary chalcogenides Cu(Cd, Zn)$_2$InTe$_4$

Crystalline semiconductors with intrinsically low lattice thermal conductivity ($\mathcal{K}$) are vital for device applications such as barrier coatings and thermoelectrics. Quaternary chalcogenide semiconductors such as CuCd$_2$InTe$_4$ and CuZn$_2$InTe$_4$ are experimentally shown to exhibit low $\mathcal{K}$, yet its microscopic origin remains poorly understood. Here, we analyse their thermal transport mechanisms using a unified first-principles framework that captures both the Peierls (particle-like propagation, $\mathcal{K}_P$) and coherence (wave-like tunneling, $\mathcal{K}_C$) mechanisms of phonon transport. We show that extended antibonding states below the Fermi level lead to enhanced phonon anharmonicity and strong scattering of heat-carrying phonon modes, suppressing $\mathcal{K}$ in these chalcogenides. We show that $\mathcal{K}_P$ dominates the total thermal conductivity, while $\mathcal{K}_C$ remains negligible even under strong anharmonicity of the phonon modes. The heavier Cd ions in CuCd$_2$InTe$_4$ induce greater acoustic-optical phonon overlap and scattering compared to CuZn$_2$InTe$_4$, further lowering thermal conductivity of the former. Additionally, grain boundary scattering in realistic samples contributes to further suppression of thermal transport. Our findings establish the atomistic origins of low $\mathcal{K}$ in quaternary chalcogenides and offer guiding principles for designing low-thermal-conductivity semiconductors.

cond-mat.mtrl-sci↗

arXiv2025

FlexDoc: Parameterized Sampling for Diverse Multilingual Synthetic Documents for Training Document Understanding Models

Developing document understanding models at enterprise scale requires large, diverse, and well-annotated datasets spanning a wide range of document types. However, collecting such data is prohibitively expensive due to privacy constraints, legal restrictions, and the sheer volume of manual annotation needed - costs that can scale into millions of dollars. We introduce FlexDoc, a scalable synthetic data generation framework that combines Stochastic Schemas and Parameterized Sampling to produce realistic, multilingual semi-structured documents with rich annotations. By probabilistically modeling layout patterns, visual structure, and content variability, FlexDoc enables the controlled generation of diverse document variants at scale. Experiments on Key Information Extraction (KIE) tasks demonstrate that FlexDoc-generated data improves the absolute F1 Score by up to 11% when used to augment real datasets, while reducing annotation effort by over 90% compared to traditional hard-template methods. The solution is in active deployment, where it has accelerated the development of enterprise-grade document understanding models while significantly reducing data acquisition and annotation costs.

cs.AI↗

arXiv2025

RCI: A Score for Evaluating Global and Local Reasoning in Multimodal Benchmarks

Multimodal Large Language Models (MLLMs) have achieved impressive results on vision-language benchmarks, yet it remains unclear whether these benchmarks assess genuine global reasoning or allow success via localized visual cues. Existing evaluation methods do not explicitly measure this distinction, hindering effective dataset curation and real-world focused model development. We introduce Region Comprehension Index (RCI), the first model-based score to directly quantify a dataset's reliance on global versus local visual information. RCI systematically compares reference-model performance on image patches versus full images, revealing if tasks require holistic image understanding or can be solved with partial or localized visual cues. When applying RCI to 13 widely used multimodal benchmarks, we observed that most of them favor localized reasoning and exhibit significant spatial biases, indicating potential risks in real-world applications. RCI equips researchers & practitioners with an actionable tool for diagnosing & mitigating these biases, enabling the construction of datasets and benchmarks to foster the development of robust, enterprise-ready multimodal systems.

cs.CV↗

arXiv2025

PCRI: Measuring Context Robustness in Multimodal Models for Enterprise Applications

The reliability of Multimodal Large Language Models (MLLMs) in real-world settings is often undermined by sensitivity to irrelevant or distracting visual context, an aspect not captured by existing evaluation metrics. We introduce the \textbf{Patch Context Robustness Index (PCRI)}, the first systematic and interpretable score for quantifying MLLM robustness to variations in visual context granularity, measuring performance changes between localized image patches and full-image input. Applying PCRI to 19 state-of-the-art MLLMs across 15 vision-language benchmarks, we find that most leading models remain brittle to background noise, with only a few, such as InternVL2-26B and Qwen2VL-72B, demonstrating consistent robustness across tasks. PCRI analysis also highlights how different model architectures handle and integrate visual context, offering actionable diagnostic insight for both researchers and practitioners. PCRI enables rigorous comparison of context robustness, supporting principled model selection and guiding the development of future architectures and training strategies for robust, real-world deployment.

cs.CV↗

arXiv2025

Quantum metric induced nonlinear thermal noise in PT-symmetric antiferromagnets

Electrical current fluctuations provide a powerful and unconventional probe of band geometry in quantum materials. In particular, intrinsic noise components that are independent of the relaxation time reveal universal band-geometric properties of Bloch electrons. Here, we identify a distinct intrinsic contribution to current noise at second order in the electric field, which is governed by the quantum metric. This effect arises from field-induced modifications to Bloch wavefunctions and band energies, and it dominates in PT-symmetric antiferromagnets where Berry curvature based contributions are forbidden by symmetry. Applying our theory to CuMnAs, we demonstrate that thermal noise provides a direct signature of quantum metric in PT-symmetric antiferromagnets.

cond-mat.mes-hall↗

arXiv2025

Spin band geometry drives intrinsic thermal spin magnetization and current

Generating spin magnetization and spin currents without magnetic or electric fields is a key frontier in spin caloritronics. Spin responses driven by thermal gradients offer a promising route, though the band geometric origin of intrinsic mechanisms, especially in non-magnetic materials, remains poorly understood. Here we develop a unified quantum theory of thermal spin magnetization and spin currents in itinerant electrons, rooted in spin band geometry with both Fermi-surface and Fermi-sea contributions. We identify two key geometric quantities: the spin-velocity metric tensor, which governs thermal spin magnetization, and the spin geometric tensor, combining spin Berry curvature and spin quantum metric, which generates thermal spin currents. These intrinsic contributions persist and can even dominate in non-magnetic insulators. Numerical calculations for chiral metal RhGe and antiferromagnet CuMnAs demonstrate sizable thermal spin responses near band crossings. Our results establish the band geometric origin of thermal spin transport and provide guiding principles for discovering and engineering next-generation spin caloritronic materials.

cond-mat.mes-hall↗

arXiv2025

AccessEval: Benchmarking Disability Bias in Large Language Models

Large Language Models (LLMs) are increasingly deployed across diverse domains but often exhibit disparities in how they handle real-life queries. To systematically investigate these effects within various disability contexts, we introduce \textbf{AccessEval (Accessibility Evaluation)}, a benchmark evaluating 21 closed- and open-source LLMs across 6 real-world domains and 9 disability types using paired Neutral and Disability-Aware Queries. We evaluated model outputs with metrics for sentiment, social perception, and factual accuracy. Our analysis reveals that responses to disability-aware queries tend to have a more negative tone, increased stereotyping, and higher factual error compared to neutral queries. These effects show notable variation by domain and disability type, with disabilities affecting hearing, speech, and mobility disproportionately impacted. These disparities reflect persistent forms of ableism embedded in model behavior. By examining model performance in real-world decision-making contexts, we better illuminate how such biases can translate into tangible harms for disabled users. This framing helps bridges the gap between technical evaluation and user impact, reinforcing importance of bias mitigation in day-to-day applications. Our dataset is publicly available at: https://huggingface.co/datasets/Srikant86/AccessEval

cs.CL↗

arXiv2025

Band geometric transverse current driven by inhomogeneous AC electric field

We develop a semiclassical theory for electron wavepacket dynamics in the presence of an inhomogeneous AC electric field. While static electric-field gradients are known to generate charge transport governed by the quantum metric, we show that AC field gradients induce an additional geometric current that vanishes in the DC limit. This response originates from a novel band-geometric quantity, the higher-order connection (HOC) tensor, constructed from cubic products of interband Berry connections. We derive explicit expressions for the AC current and identify the symmetry conditions under which it arises. Remarkably, inhomogeneous AC fields can generate an anomalous Hall-like response even in nonmagnetic systems. Applying the theory to Bernal-stacked bilayer graphene, we demonstrate that the HOC-induced response produces a measurable Hall current peaking at band edges. These results establish inhomogeneous AC fields as a powerful probe of higher-order band geometric quantities beyond Berry curvature and the quantum metric.

cond-mat.mes-hall↗

arXiv2025

Disorder-driven Weyl-Kondo Semimetal Phase in WTe$_2$

In this Letter, we report the observation of disorder-driven anisotropic Kondo screening and spontaneous Hall effect in bulk WTe${_2}$, a nonmagnetic type-II Weyl semimetal. We show that Kondo scattering emerges more prominently in disordered samples and produces magnetoresistance that is strongly anisotropic with respect to both current and magnetic field orientation, reflecting the underlying type-II Weyl dispersion. Strikingly, we find a spontaneous Hall effect in zero magnetic field, whose magnitude is enhanced with disorder, together with a large second-harmonic Hall signal exhibiting quadratic current scaling. Our analysis indicates that disorder-driven Kondo interactions pin the Fermi level near the Weyl nodes. This enhances the Berry curvature-driven nonequilibrium transport, accounting for both the second-order and spontaneous Hall responses. These findings establish disordered WTe${_2}$ as a platform hosting Weyl-Kondo fermions and highlight disorder as an effective control knob for inducing correlated topological phases in weakly correlated Weyl semimetals.

cond-mat.mes-hall↗

arXiv2025

Band Geometry Induced Third-Harmonic Generation

Third-harmonic generation (THG) is a key nonlinear optical process for ultrafast imaging, terahertz (THz) signal generation, and symmetry-sensitive probes, often dominating in centrosymmetric materials where lower-order responses vanish. Yet, the role of band geometry, Fermi surface effects, and disorder in enabling large and tunable THG remains poorly understood. Here, we develop a finite-frequency quantum kinetic theory of THG based on the density matrix formalism, deriving the third-harmonic conductivity tensor. Our framework isolates five distinct band-geometric contributions to interband and intraband processes, separates Fermi sea from Fermi surface terms, and incorporates disorder effects phenomenologically. We further provide a complete symmetry classification of THG for all 122 magnetic point groups. Applying the theory to the spin-split altermagnet RuO$_2$, we trace its THG response to specific geometric terms. These results establish a predictive foundation for designing materials with enhanced and tunable THG in the finite-frequency regime.

physics.optics↗

arXiv2025

High-Throughput Screening of 2D Photocatalyst Heterostructures with Suppressed Electron-Hole Recombination for Solar Water Splitting

Efficient and scalable photocatalysts for solar water splitting remain a critical challenge in renewable energy research. The work presents a high-throughput first-principles discovery of two-dimensional (2D) type-II van der Waals heterostructures (vdWHs) optimized for visible-light-driven photocatalytic water splitting. We screened 482 heterostructures constructed from 60 experimentally realizable 2D monolayers and identified 148 stable type-II vdWHs with spatially separated valence and conduction band edges, out of which 65 satisfy the thermodynamic redox conditions for water splitting over a broad pH range. Among these, the best two, MoTe2/Tl2O and MoSe2/WSe2, exhibit a high visible-light absorption coefficient exceeding 0.6X10^6 cm-1, resulting in a high power conversion efficiency of 2%. Quantum kinetic analysis of the hydrogen evolution reaction (HER) reveals nearly barrierless free energy profiles across multiple adsorption sites. Our study further reveals that intrinsic interlayer electric fields in these vdWHs drive directional charge separation, suppressing carrier recombination. Our results establish a design framework for using type-II 2D heterostructures as tunable and experimentally accessible 2D photocatalysts for efficient hydrogen production.

cond-mat.mtrl-sci↗

arXiv2025

Fourier Domain Adaptation for Traffic Light Detection in Adverse Weather

Traffic light detection under adverse weather conditions remains largely unexplored in ADAS systems, with existing approaches relying on complex deep learning methods that introduce significant computational overheads during training and deployment. This paper proposes Fourier Domain Adaptation (FDA), which requires only training data modifications without architectural changes, enabling effective adaptation to rainy and foggy conditions. FDA minimizes the domain gap between source and target domains, creating a dataset for reliable performance under adverse weather. The source domain merged LISA and S2TLD datasets, processed to address class imbalance. Established methods simulated rainy and foggy scenarios to form the target domain. Semi-Supervised Learning (SSL) techniques were explored to leverage data more effectively, addressing the shortage of comprehensive datasets and poor performance of state-of-the-art models under hostile weather. Experimental results show FDA-augmented models outperform baseline models across mAP50, mAP50-95, Precision, and Recall metrics. YOLOv8 achieved a 12.25% average increase across all metrics. Average improvements of 7.69% in Precision, 19.91% in Recall, 15.85% in mAP50, and 23.81% in mAP50-95 were observed across all models, demonstrating FDA's effectiveness in mitigating adverse weather impact. These improvements enable real-world applications requiring reliable performance in challenging environmental conditions.

cs.CV↗

arXiv2025

Layered Topological Antiferromagnetic Metal at Room Temperature -- YbMn$_2$Ge$_2$

Metallic antiferromagnets are essential for efficient spintronic applications due to their fast switching and high mobility, yet room-temperature metallic antiferromagnets are rare. Here, we investigate YbMn$_2$Ge$_2$, a room temperature antiferromagnet, and establish it as an exfoliable layered metal with altermagnetic surface states. Using multi-orbital Hubbard model calculations, we reveal that its robust metallic AFM ordering is stabilized by electronic correlations and a partially nested Fermi surface. Furthermore, we show that YbMn$_2$Ge$_2$ hosts symmetry-protected topological Dirac crossings, connecting unique even-order spin-polarized surface states with parabolic and inverted Mexican-hat-like dispersion. Our findings position YbMn$_2$Ge$_2$ as a promising platform for exploring the interplay of correlation, topology, and surface altermagnetism of layered antiferromagnets.

cond-mat.mtrl-sci↗