arXiv subjects

Anand Kumar

Publications and source records attributed to Anand Kumar.

At least 19 recordsLinked to original sources

arXiv2026

Symmetry-resolved tree tensor network analysis of Bell-state discrimination with ancilla-assisted passive linear optics

Ancillary photons allow passive linear-optical Bell-state discrimination to exceed the one-half probability of success limit achievable with vacuum auxiliary modes. For a fixed analyzer and ancillary resource, the discrimination probability is determined by the photon-counting patterns that occur uniquely for each Bell input. We develop a photon-number-resolved tree tensor network method for evaluating these detector supports in recursively structured analyzers. Photon-number conservation and mode symmetries separate most of the Bell-state outputs before the remaining detector support is examined. We establish a tree tensor network based construction to evaluate the remaining support without need of recalling all relevant photon detector patterns separately. We apply the method to recursive, product, and asymmetric ancillary states with up to $32$ optical modes and reproduce known analytical and literature benchmarks. For factorized ancillary resources with fixed photon numbers in halves of the analyzer and the symmetry property used in our analysis, we also derive an exact composition relation; for reflected asymmetric pairings the success probability is the arithmetic mean of the corresponding symmetric configurations. The complete enumeration for smaller systems, permanent-based amplitude calculations, and independent evaluations provide additional checks of the numerical results.

quant-ph↗

arXiv2026

Uncertainty-guided active learning for surrogate prediction of stream-finishing wear fields

In stream finishing, the wear experienced by a workpiece depends strongly on its orientation within the rotating abrasive media. Determining suitable orientations to achieve uniform wear requires evaluating the wear-rate field over all feasible orientations. Although the discrete element method (DEM) accurately resolves particle interactions, simulating hundreds of feasible orientations for a new geometry is computationally expensive. We present an uncertainty-guided surrogate framework that predicts, directly from geometry, the three fields governing erosion: per-triangle normal impact velocity, tangential impact velocity, and particle impact flux. These fields are combined through the Finnie wear model to reconstruct the wear-rate distribution. The surrogate employs a deep ensemble whose disagreement estimates epistemic uncertainty, enabling an active-learning strategy that selectively performs DEM simulations for the most uncertain orientations. Trained using only $13\%$ of the $696$ feasible orientations, the surrogate achieves Spearman rank correlations of $0.93$, $0.89$, and $0.93$ for the normal impact velocity, tangential impact velocity, and particle impact flux, respectively. Moreover, the predicted uncertainty is well calibrated, reliably anticipating prediction error and the fidelity of the reconstructed wear field, which matches DEM with a Spearman rank correlation of up to $0.97$ for low-uncertainty orientations and degrades in a controlled manner as uncertainty increases.

cond-mat.mtrl-sci↗

arXiv2026

Insights into the Nature of Quantum Emitters in Electron-Irradiated hexagonal Boron Nitride

Quantum emitters in hexagonal boron nitride (hBN) have emerged as a promising solid-state platform for quantum technology applications. However, a persistent challenge in the field is the unclear origin of many observed emission lines, particularly in the visible range, which can be difficult to distinguish from signals arising from organic or process-induced contamination during sample preparations and handling. This ambiguity limits both the reproducibility of emitter generation and the reliable identification of truly intrinsic quantum defects. This work provides a step-by-step framework to assess whether quantum emitters in electron-irradiated hBN are associated with organic contaminants introduced during sample preparation. We employ hyperspectral imaging, thermal annealing, and oxygen plasma etching to investigate the origin of the green-yellow emitters in electron-irradiated hBN. The combined results not only rule out organic contamination as the source of emission but also provide insight into the spectral variability, thermal stability, and vertical localization of the emitters generated in electron-irradiated hBN that was created without any pre- or post-processing. In addition, our experiments demonstrate the feasibility of creating stable emitters in hBN with thicknesses below 10 nm. These findings provide practical guidance for the identification and controlled implementation of hBN-based single-photon emitters in quantum photonic devices.

physics.optics↗

arXiv2026

$\texttt{YC-Bench}$: Benchmarking AI Agents for Long-Term Planning and Consistent Execution

As LLM agents tackle increasingly complex tasks, a critical question is whether they can maintain strategic coherence over long horizons: planning under uncertainty, learning from delayed feedback, and adapting when early mistakes compound. We introduce $\texttt{YC-Bench}$, a benchmark that evaluates these capabilities by tasking an agent with running a simulated startup over a one-year horizon spanning hundreds of turns. The agent must manage employees, select task contracts, and maintain profitability in a partially observable environment where adversarial clients and growing payroll create compounding consequences for poor decisions. We evaluate 12 models, both proprietary and open source, across 3 seeds each. Only three models consistently surpass the starting capital of \$200K, with Claude Opus 4.6 achieving the highest average final funds at \$1.27 M, followed by GLM-5 at \$1.21 M at 11$\times$ lower inference cost. Scratchpad usage, the sole mechanism for persisting information across context truncation, is the strongest predictor of success, and adversarial client detection is the primary failure mode, accounting for $47\%$ of bankruptcies. Our analysis reveals that frontier models still fail through distinct failure modes such as over-parallelization, demonstrating the capability gaps for long-horizon performance. $\texttt{YC-Bench}$ is open-source, reproducible, and configurable.

cs.CL↗

arXiv2025

The Valley of Code Reasoning: Scaling Knowledge Distillation of Large Language Models

Distilling the thinking traces of a Large Language Model (LLM) with reasoning capabilities into a smaller model has been proven effective. Yet, there is a scarcity of work done on how model performances scale with the quantity of distillation data. In this work, we study the scaling trend of distilling competitive coding skills on two small non-reasoning LLMs. We validate the hypothesis that there is a $\textit{valley of code reasoning}$: downstream performance on competitive coding first drops as data quantity increases, then it steadily increases in a sharper-than-log-linear fashion. Having identified the trend, we further fine-tune the models at two different distillation stages on the same data to ground conclusions on their respective learning phases. We learn that across stages in the low and medium-low data regimes, small models benefit significantly from easier coding questions than from harder ones. We also find that, surprisingly, the correctness of outputs in training data makes no difference to distillation outcomes. Our work represents a step forward in understanding the training dynamics of code reasoning distillation outside intuition

cs.CL↗

arXiv2025

Impatient Users Confuse AI Agents: High-fidelity Simulations of Human Traits for Testing Agents

Despite rapid progress in building conversational AI agents, robustness is still largely untested. Small shifts in user behavior, such as being more impatient, incoherent, or skeptical, can cause sharp drops in agent performance, revealing how brittle current AI agents are. Today's benchmarks fail to capture this fragility: agents may perform well under standard evaluations but degrade spectacularly in more realistic and varied settings. We address this robustness testing gap by introducing TraitBasis, a lightweight, model-agnostic method for systematically stress testing AI agents. TraitBasis learns directions in activation space corresponding to steerable user traits (e.g., impatience or incoherence), which can be controlled, scaled, composed, and applied at inference time without any fine-tuning or extra data. Using TraitBasis, we extend $\tau$-Bench to $\tau$-Trait, where user behaviors are altered via controlled trait vectors. We observe on average a 2%-30% performance degradation on $\tau$-Trait across frontier models, highlighting the lack of robustness of current AI agents to variations in user behavior. Together, these results highlight both the critical role of robustness testing and the promise of TraitBasis as a simple, data-efficient, and compositional tool. By powering simulation-driven stress tests and training loops, TraitBasis opens the door to building AI agents that remain reliable in the unpredictable dynamics of real-world human interactions. We have open-sourced $\tau$-Trai across four domains: airline, retail, telecom, and telehealth, so the community can systematically QA their agents under realistic, behaviorally diverse intents and trait scenarios: https://github.com/collinear-ai/tau-trait.

cs.AI↗

arXiv2025

Enhancement in phase sensitivity in displacement-assisted SU(1,1) interferometer via photon recycling

We propose a novel method for enhancing phase estimation in the displacement-assisted SU(1,1) (DSU(1,1)) interferometer by incorporating the photon recycling technique, evaluated under single-intensity detection (SID) and homodyne detection (HD) schemes. Our analysis showed that utilizing the photon recycling technique, the photon-recycled DSU(1,1) interferometer performs better than the conventional DSU(1,1) interferometer for some conditions. We also showed that this improvement is possible in both SID and HD schemes. In addition, to discuss the maximum sensitivity achieved by our proposed model, we have calculated the quantum Cram\'{e}r-Rao bound (QCRB) within the framework and found that our proposed model approaches the QCRB. Therefore, we believe that our findings offer a promising new approach to improve phase sensitivity through photon recycling.

quant-ph↗

arXiv2025

Intuition to Evidence: Measuring AI's True Impact on Developer Productivity

We present a comprehensive real-world evaluation of AI-assisted software development tools deployed at enterprise scale. Over one year, 300 engineers across multiple teams integrated an in-house AI platform (DeputyDev) that combines code generation and automated review capabilities into their daily workflows. Through rigorous cohort analysis, our study demonstrates statistically significant productivity improvements, including an overall 31.8% reduction in PR review cycle time. Developer adoption was strong, with 85% satisfaction for code review features and 93% expressing a desire to continue using the platform. Adoption patterns showed systematic scaling from 4% engagement in month 1 to 83% peak usage by month 6, stabilizing at 60% active engagement. Top adopters achieved a 61% increase in code volume pushed to production, contributing to approximately 30 to 40% of code shipped to production through this tool, accounting for an overall 28% increase in code shipment volume. Unlike controlled benchmark evaluations, our longitudinal analysis provides empirical evidence from production environments, revealing both the transformative potential and practical deployment challenges of integrating AI into enterprise software development workflows.

cs.SE↗

arXiv2025

Mobile-Friendly Deep Learning for Plant Disease Detection: A Lightweight CNN Benchmark Across 101 Classes of 33 Crops

Plant diseases are a major threat to food security globally. It is important to develop early detection systems which can accurately detect. The advancement in computer vision techniques has the potential to solve this challenge. We have developed a mobile-friendly solution which can accurately classify 101 plant diseases across 33 crops. We built a comprehensive dataset by combining different datasets, Plant Doc, PlantVillage, and PlantWild, all of which are for the same purpose. We evaluated performance across several lightweight architectures - MobileNetV2, MobileNetV3, MobileNetV3-Large, and EfficientNet-B0, B1 - specifically chosen for their efficiency on resource-constrained devices. The results were promising, with EfficientNet-B1 delivering our best performance at 94.7% classification accuracy. This architecture struck an optimal balance between accuracy and computational efficiency, making it well-suited for real-world deployment on mobile devices.

cs.CV↗

arXiv2025

DeputyDev -- AI Powered Developer Assistant: Breaking the Code Review Logjam through Contextual AI to Boost Developer Productivity

This study investigates the implementation and efficacy of DeputyDev, an AI-powered code review assistant developed to address inefficiencies in the software development process. The process of code review is highly inefficient for several reasons, such as it being a time-consuming process, inconsistent feedback, and review quality not being at par most of the time. Using our telemetry data, we observed that at TATA 1mg, pull request (PR) processing exhibits significant inefficiencies, with average pick-up and review times of 73 and 82 hours, respectively, resulting in a 6.2 day closure cycle. The review cycle was marked by prolonged iterative communication between the reviewing and submitting parties. Research from the University of California, Irvine indicates that interruptions can lead to an average of 23 minutes of lost focus, critically affecting code quality and timely delivery. To address these challenges, we developed DeputyDev's PR review capabilities by providing automated, contextual code reviews. We conducted a rigorous double-controlled A/B experiment involving over 200 engineers to evaluate DeputyDev's impact on review times. The results demonstrated a statistically significant reduction in both average per PR (23.09%) and average per-line-of-code (40.13%) review durations. After implementing safeguards to exclude outliers, DeputyDev has been effectively rolled out across the entire organisation. Additionally, it has been made available to external companies as a Software-as-a-Service (SaaS) solution, currently supporting the daily work of numerous engineering professionals. This study explores the implementation and effectiveness of AI-assisted code reviews in improving development workflow timelines and code.

cs.SE↗

arXiv2025

SparseEMG: Computational Design of Sparse EMG Layouts for Sensing Gestures

Gesture recognition with electromyography (EMG) is a complex problem influenced by gesture sets, electrode count and placement, and machine learning parameters (e.g., features, classifiers). Most existing toolkits focus on streamlining model development but overlook the impact of electrode selection on classification accuracy. In this work, we present the first data-driven analysis of how electrode selection and classifier choice affect both accuracy and sparsity. Through a systematic evaluation of 28 combinations (4 selection schemes, 7 classifiers), across six datasets, we identify an approach that minimizes electrode count without compromising accuracy. The results show that Permutation Importance (selection scheme) with Random Forest (classifier) reduces the number of electrodes by 53.5\%. Based on these findings, we introduce SparseEMG, a design tool that generates sparse electrode layouts based on user-selected gesture sets, electrode constraints, and ML parameters while also predicting classification performance. SparseEMG supports 50+ unique gestures and is validated in three real-world applications using different hardware setups. Results from our multi-dataset evaluation show that the layouts generated from the SparseEMG design tool are transferable across users with only minimal variation in gesture recognition performance.

cs.HC↗

arXiv2025

Temperature-Dependent Emission Spectroscopy of Quantum Emitters in Hexagonal Boron Nitride

Color centers in hexagonal boron nitride (hBN) have attracted significant interest due to their potential applications in future optical quantum technologies. For most applications, scalable on-demand fabrication is a key requirement. Recent advances using localized electron irradiation have demonstrated near-identical emitters in the blue and yellow spectral regions. While the blue emitters have been demonstrated in cryogenic temperatures, the yellow emitters remain uncharacterized under such conditions. In this work, we therefore extended the study of yellow emitters to cryogenic temperatures. Initially, multiple spectral features were observed, prompting a systematic investigation that led to the identification of a defect emission centered around 547.5 nm with high brightness and excellent photostability. By tuning the excitation wavelength, we are able to distinguish Raman scattering peaks from the emitter emission. Further analysis of the vibronic emissions allowed us to identify an optical phonon mode, whose contribution becomes increasingly dominant at elevated temperatures. Photoluminescence excitation spectroscopy (PLE) reveals excitation through this phonon mode enhances the emission by almost 5-fold in cryogenic temperature. Temperature-dependent studies further elucidate the role of phonons in the emission process. These observations deepen our understanding of the nature of the emitters, opening new avenues for precise tuning of quantum light sources.

physics.optics↗

arXiv2025

All-dry pick-up and transfer method for quantum emitter arrays in hexagonal boron nitride

Single photon emitters in hexagonal boron nitride are based on fluorescent point-like defects. These defects typically have exceptional photophysical properties and therefore been the focus of extensive research due to their potential to advance photonic quantum technologies. However, achieving scalable integration of these emitters to arbitrary platforms with high yield while retaining their characteristics remains a significant challenge, particularly when the target substrate is not compatible with the fabrication method. In this work, we introduce an all-dry transfer method aimed at addressing these challenges with improved effectiveness compared to existing techniques. This polymer stamp-assisted transfer method maintains high output and preserves the fundamental characteristics of the emitters while eliminating wet chemical processes. A comprehensive post-transfer characterization verified not only the maintenance of the defining characteristic of a single photon emitter, the second-order correlation function $g^{(2)}(0)$, but also showed improvement by about 46%. In contrast, the lifetime, emission spectrum, and the photostability showed only negligible change, demonstrating that the characteristics of the emitters were retained during the transfer process. This transfer technique has success rate of 81.8%, determined by the proportion of single photon emitters that retain their optical and preserve physical structure post-transfer. This high success rate shows the potential to scale the integration of single photon emitters across diverse platforms. We expect that this process contributes to the applications of boron nitride defects in quantum technologies.

physics.optics↗

arXiv2025

GS-TransUNet: Integrated 2D Gaussian Splatting and Transformer UNet for Accurate Skin Lesion Analysis

We can achieve fast and consistent early skin cancer detection with recent developments in computer vision and deep learning techniques. However, the existing skin lesion segmentation and classification prediction models run independently, thus missing potential efficiencies from their integrated execution. To unify skin lesion analysis, our paper presents the Gaussian Splatting - Transformer UNet (GS-TransUNet), a novel approach that synergistically combines 2D Gaussian splatting with the Transformer UNet architecture for automated skin cancer diagnosis. Our unified deep learning model efficiently delivers dual-function skin lesion classification and segmentation for clinical diagnosis. Evaluated on ISIC-2017 and PH2 datasets, our network demonstrates superior performance compared to existing state-of-the-art models across multiple metrics through 5-fold cross-validation. Our findings illustrate significant advancements in the precision of segmentation and classification. This integration sets new benchmarks in the field and highlights the potential for further research into multi-task medical image analysis methodologies, promising enhancements in automated diagnostic systems.

cs.CV↗

arXiv2024

IntroStyle: Training-Free Introspective Style Attribution using Diffusion Features

Text-to-image (T2I) models have recently gained widespread adoption. This has spurred concerns about safeguarding intellectual property rights and an increasing demand for mechanisms that prevent the generation of specific artistic styles. Existing methods for style extraction typically necessitate the collection of custom datasets and the training of specialized models. This, however, is resource-intensive, time-consuming, and often impractical for real-time applications. We present a novel, training-free framework to solve the style attribution problem, using the features produced by a diffusion model alone, without any external modules or retraining. This is denoted as Introspective Style attribution (IntroStyle) and is shown to have superior performance to state-of-the-art models for style attribution. We also introduce a synthetic dataset of Artistic Style Split (ArtSplit) to isolate artistic style and evaluate fine-grained style attribution performance. Our experimental results on WikiArt and DomainNet datasets show that \ours is robust to the dynamic nature of artistic styles, outperforming existing methods by a wide margin.

cs.CV↗

arXiv2024

Quantitative investigation of quantum emitter yield in drop-casted hexagonal boron nitride nanoflakes

Single photon emitters (SPEs) are a key component for their use as pure photon source in quantum technologies. In this study, we investigate the generation of SPEs from drop-casted hexagonal boron nitride (hBN) nanoflakes, examining the influence of the immersion solution and the source of hBN. We show that, depending on the utilized supplier and solution the number and quality of the emitters changes. We perform a comprehensive optical characterization of the deposited nanoflakes to assess the quality of the generated SPEs. We show quantitative data on SPE yields, highlighting significant variations among solvents and different sources of hBN. This holds particular significance for employing drop-casted nanoflakes as SPE sources in quantum communication, sensing, and imaging. Our method is easily expandable to all kinds of surfaces and can be done without requiring complex fabrication steps and equipment, thus providing the necessary scalability required for industrial quantum applications.

physics.optics↗

arXiv2023

Comparative study of quantum emitter fabrication in wide bandgap materials using localized electron irradiation

Quantum light sources are crucial foundational components for various quantum technology applications. With the rapid development of quantum technology, there has been a growing demand for materials with the capability of hosting quantum emitters. One such material platform uses fluorescent defects in hexagonal boron nitride (hBN) that can host deep sublevels within the bandgap. The localized electron irradiation has shown its effectiveness in generating deep sublevels to induce single emitters in hBN. The question is whether localized (electron beam) irradiation is a reliable tool for creating emitters in other wide bandgap materials and its uniqueness to hBN. Here, we investigate and compare the fabrication of quantum emitters in hBN and exfoliated muscovite mica flakes along with other 3D crystals, such as silicon carbide and gallium nitride, which are known to host quantum emitters. We used our primary fabrication technique of localized electron irradiation using a standard scanning electron microscope. To complement our experimental work, we employed density functional theory simulations to study the atomic structures of defects in mica. While our fabrication technique allows one to create hBN quantum emitters with a high yield and high single photon purity, it is unable to fabricate single emitters in the other solid-state crystals under investigation. This allows us to draw conclusions on the emitter fabrication mechanism in hBN, which could rely on activating pre-existing defects by charge state manipulation. Therefore, we provide an essential step toward the identification of hBN emitters and their formation process.

physics.app-ph↗

arXiv2023

Polarization dynamics of solid-state quantum emitters

Quantum emitters in solid-state crystals have recently attracted a lot of attention due to their simple applicability in optical quantum technologies. The polarization of single photons generated by quantum emitters is one of the key parameters that play a crucial role in the applications, such as quantum computation that uses the indistinguishability of photons. However, the degree of single photon polarization is typically quantified using time-averaged photoluminescence intensity of single emitters, which provides limited information about the dipole properties in solids. In this work, we use single defects in hexagonal boron nitride and nanodiamond as efficient room-temperature single photon sources to reveal the origin and the temporal evolution of dipole orientation in solid-state quantum emitters. The angle of excitation and emission dipoles relative to the crystal axes are determined experimentally and then calculated using density functional theory, which results in characteristic angles for every specific defect that can be used as an efficient tool for defect identification and understanding their atomic structure. Moreover, the temporal polarization dynamics reveal a strongly modified linear polarization visibility that depends on the excited state decay time of individual excitation. This effect can be traced back potentially to the excitation of excess charges in the local crystal environment. Understanding such hidden time-dependent mechanisms can further be used to improve the performance of polarization-sensitive experiments, in particular that of quantum communication with single photon emitters.

quant-ph↗