arXiv subjects

Sachin Yadav

Publications and source records attributed to Sachin Yadav.

At least 19 recordsLinked to original sources

arXiv2026

Extreme Meta-Classification for Large-Scale Zero-Shot Retrieval

We develop accurate and efficient solutions for large-scale retrieval tasks where novel (zero-shot) items can arrive continuously at a rapid pace. Conventional Siamese-style approaches embed both queries and items through a small encoder and retrieve the items lying closest to the query. While this approach allows efficient addition and retrieval of novel items, the small encoder lacks sufficient capacity for the necessary world knowledge in complex retrieval tasks. The extreme classification approaches have addressed this by learning a separate classifier for each item observed in the training set which significantly increases the representation capacity of the model. Such classifiers outperform Siamese approaches on observed items, but cannot be trained for novel items due to data and latency constraints. To bridge these gaps, this paper develops: (1) A new algorithmic framework, EMMETT, which efficiently synthesizes classifiers on-the-fly for novel items, by relying on the readily available classifiers for observed items; (2) A new algorithm, IRENE, which is a simple and effective instance of EMMETT that is specifically suited for large-scale deployments, and (3) A new theoretical framework for analyzing the generalization performance in large-scale zero-shot retrieval which guides our algorithm and training related design decisions. Comprehensive experiments are conducted on a wide range of retrieval tasks which demonstrate that IRENE improves the zero-shot retrieval accuracy by up to 15% points in Recall@10 when added on top of leading encoders. Additionally, on an online A/B test in a large-scale ad retrieval task in a major search engine, IRENE improved the ad click-through rate by 4.2%. Lastly, we validate our design choices through extensive ablative experiments. The source code for IRENE is available at https://aka.ms/irene.

cs.IR↗

arXiv2026

The LSCD Benchmark: a Testbed for Diachronic Word Meaning Tasks

Lexical Semantic Change Detection (LSCD) is a complex, lemma-level task, which is usually operationalized based on two subsequently applied usage-level tasks: First, Word-in-Context (WiC) labels are derived for pairs of usages. Then, these labels are represented in a graph on which Word Sense Induction (WSI) is applied to derive sense clusters. Finally, LSCD labels are derived by comparing sense clusters over time. This modularity is reflected in most LSCD datasets and models. It also leads to a large heterogeneity in modeling options and task definitions, which is exacerbated by a variety of dataset versions, preprocessing options and evaluation metrics. This heterogeneity makes it difficult to evaluate models under comparable conditions, to choose optimal model combinations or to reproduce results. Hence, we provide a benchmark repository standardizing LSCD evaluation. Through transparent implementation results become easily reproducible and by standardization different components can be freely combined. The repository reflects the task's modularity by allowing model evaluation for WiC, WSI and LSCD. This allows for careful evaluation of increasingly complex model components providing new ways of model optimization.

cs.CL↗

arXiv2025

XL-DURel: Finetuning Sentence Transformers for Ordinal Word-in-Context Classification

We propose XL-DURel, a finetuned, multilingual Sentence Transformer model optimized for ordinal Word-in-Context classification. We test several loss functions for regression and ranking tasks managing to outperform previous models on ordinal and binary data with a ranking objective based on angular distance in complex space. We further show that binary WiC can be treated as a special case of ordinal WiC and that optimizing models for the general ordinal task improves performance on the more specific binary task. This paves the way for a unified treatment of WiC modeling across different task formulations.

cs.CL↗

arXiv2025

Interleaved Gibbs Diffusion: Generating Discrete-Continuous Data with Implicit Constraints

We introduce Interleaved Gibbs Diffusion (IGD), a novel generative modeling framework for discrete-continuous data, focusing on problems with important, implicit and unspecified constraints in the data. Most prior works on discrete and discrete-continuous diffusion assume a factorized denoising distribution, which can hinder the modeling of strong dependencies between random variables in such problems. We empirically demonstrate a significant improvement in 3-SAT performance out of the box by switching to a Gibbs-sampling style discrete diffusion model which does not assume factorizability. Motivated by this, we introduce IGD which generalizes discrete time Gibbs sampling type Markov chain for the case of discrete-continuous generation. IGD allows for seamless integration between discrete and continuous denoisers while theoretically guaranteeing exact reversal of a suitable forward process. Further, it provides flexibility in the choice of denoisers, allows conditional generation via state-space doubling and inference time refinement. Empirical evaluations on three challenging generation tasks - molecule structures, layouts and tabular data - demonstrate state-of-the-art performance. Notably, IGD achieves state-of-the-art results without relying on domain-specific inductive biases like equivariant diffusion or auxiliary losses. We explore a wide range of modeling, and interleaving strategies along with hyperparameters in each of these problems.

cs.LG↗

arXiv2024

Semantic Retrieval at Walmart

In product search, the retrieval of candidate products before re-ranking is more critical and challenging than other search like web search, especially for tail queries, which have a complex and specific search intent. In this paper, we present a hybrid system for e-commerce search deployed at Walmart that combines traditional inverted index and embedding-based neural retrieval to better answer user tail queries. Our system significantly improved the relevance of the search engine, measured by both offline and online evaluations. The improvements were achieved through a combination of different approaches. We present a new technique to train the neural model at scale. and describe how the system was deployed in production with little impact on response time. We highlight multiple learnings and practical tricks that were used in the deployment of this system.

cs.IR↗

arXiv2024

Interplay between superconducting fluctuations and weak localization in disordered TiN thin films

The interplay between superconducting fluctuations (SFs) and weak localization (WL) has been probed by temperature dependent resistance [R(T)] and magnetoresistance (MR) measurements in two-dimensional disordered superconducting TiN thin films. Within a narrow band of temperature above the transition temperature Tc, the coexistence of SFs-mediated positive MR and WL-led negative MR in different range of magnetic field, as well as a crossover from positive to negative MR with increasing temperature are reported here. The crossover temperature coincides with a characteristic temperature (Tmax) at which a resistance peak appears in the zero-field R(T). The resistance peak and the associated magnetoresistance anomalies are addressed by using the quantum corrections to the conductivity (QCC) theory. We show that WL can be accounted for the observed negative MR. By introducing individual coefficients to both SFs and WL contributions, the dominance of one over the other is monitored with respect to temperature. It is observed that just above the Tc, SF dominates and with increasing temperature, the contributions from the both become comparable and finally, at Tmax, WL takes over completely. The presented approach may be adopted to compare various quantum contributions in two-dimensional superconductors particularly in the regime where both SFs and WL are pronounced.

cond-mat.supr-con↗

arXiv2024

On the Necessity of World Knowledge for Mitigating Missing Labels in Extreme Classification

Extreme Classification (XC) aims to map a query to the most relevant documents from a very large document set. XC algorithms used in real-world applications learn this mapping from datasets curated from implicit feedback, such as user clicks. However, these datasets inevitably suffer from missing labels. In this work, we observe that systematic missing labels lead to missing knowledge, which is critical for accurately modelling relevance between queries and documents. We formally show that this absence of knowledge cannot be recovered using existing methods such as propensity weighting and data imputation strategies that solely rely on the training dataset. While LLMs provide an attractive solution to augment the missing knowledge, leveraging them in applications with low latency requirements and large document sets is challenging. To incorporate missing knowledge at scale, we propose SKIM (Scalable Knowledge Infusion for Missing Labels), an algorithm that leverages a combination of small LM and abundant unstructured meta-data to effectively mitigate the missing label problem. We show the efficacy of our method on large-scale public datasets through exhaustive unbiased evaluation ranging from human annotations to simulations inspired from industrial settings. SKIM outperforms existing methods on Recall@100 by more than 10 absolute points. Additionally, SKIM scales to proprietary query-ad retrieval datasets containing 10 million documents, outperforming contemporary methods by 12% in offline evaluation and increased ad click-yield by 1.23% in an online A/B test conducted on a popular search engine. We release our code, prompts, trained XC models and finetuned SLMs at: https://github.com/bicycleman15/skim

cs.LG↗

arXiv2024

Enhancing Relevance of Embedding-based Retrieval at Walmart

Embedding-based neural retrieval (EBR) is an effective search retrieval method in product search for tackling the vocabulary gap between customer search queries and products. The initial launch of our EBR system at Walmart yielded significant gains in relevance and add-to-cart rates [1]. However, despite EBR generally retrieving more relevant products for reranking, we have observed numerous instances of relevance degradation. Enhancing retrieval performance is crucial, as it directly influences product reranking and affects the customer shopping experience. Factors contributing to these degradations include false positives/negatives in the training data and the inability to handle query misspellings. To address these issues, we present several approaches to further strengthen the capabilities of our EBR model in terms of retrieval relevance. We introduce a Relevance Reward Model (RRM) based on human relevance feedback. We utilize RRM to remove noise from the training data and distill it into our EBR model through a multi-objective loss. In addition, we present the techniques to increase the performance of our EBR model, such as typo-aware training, and semi-positive generation. The effectiveness of our EBR is demonstrated through offline relevance evaluation, online AB tests, and successful deployments to live production. [1] Alessandro Magnani, Feng Liu, Suthee Chaidaroon, Sachin Yadav, Praveen Reddy Suram, Ajit Puthenputhussery, Sijie Chen, Min Xie, Anirudh Kashi, Tony Lee, et al. 2022. Semantic retrieval at walmart. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 3495-3503.

cs.IR↗

arXiv2024

AlN/Si interface engineering to mitigate RF losses in MOCVD grown GaN-on-Si substrates

Fabrication of low-RF loss GaN-on-Si HEMT stacks is critical to enable competitive front-end-modules for 5G and 6G applications. The main contribution to RF losses is the interface between the III-N layer and the HR Si wafer, more specifically the AlN/Si interface. At this interface, a parasitic surface conduction layer exists in Si, which decreases the substrate effective resistivity sensed by overlying circuitry below the nominal Si resistivity. However, a clear understanding of this interface with control of the parasitic channel is lacking. In this letter, a detailed physical and electrical description of MOCVD-grown AlN/Si structures is presented. The presence of a $\text{SiC}_\text{x}\text{N}_\text{y}$ interfacial layer is revealed and its importance for RF losses is shown. Through C-V and I-V characterisation, an increase in the C concentration of this interfacial layer is linked to the formation of negative charge at the AlN/Si interface, which counteracts the positive charge present in the 0-predose limit. The variation of TMAl predose is shown to allow precise tuning of the C composition and, consequently, the resulting interface charge. Notably, a linear relationship between predose and net interface charge is observed and confirmed by the fabrication of an AlN/Si sample with close to zero net charge. In addition, a higher $D_{it}$ ($\sim 2\times 10^{12}$ cm$^\text{-2}$) for such compensated samples is observed and can contribute to low RF loss. An exceptionally high effective resistivity of above 8 k$Ω\cdot$cm is achieved, corresponding to an RF loss below 0.3 dB/mm at 10 GHz.

physics.app-ph↗

arXiv2024

Large Language Models for Relevance Judgment in Product Search

High relevance of retrieved and re-ranked items to the search query is the cornerstone of successful product search, yet measuring relevance of items to queries is one of the most challenging tasks in product information retrieval, and quality of product search is highly influenced by the precision and scale of available relevance-labelled data. In this paper, we present an array of techniques for leveraging Large Language Models (LLMs) for automating the relevance judgment of query-item pairs (QIPs) at scale. Using a unique dataset of multi-million QIPs, annotated by human evaluators, we test and optimize hyper parameters for finetuning billion-parameter LLMs with and without Low Rank Adaption (LoRA), as well as various modes of item attribute concatenation and prompting in LLM finetuning, and consider trade offs in item attribute inclusion for quality of relevance predictions. We demonstrate considerable improvement over baselines of prior generations of LLMs, as well as off-the-shelf models, towards relevance annotations on par with the human relevance evaluators. Our findings have immediate implications for the growing field of relevance judgment automation in product search.

cs.IR↗

arXiv2024

Towards Automating Text Annotation: A Case Study on Semantic Proximity Annotation using GPT-4

This paper explores using GPT-3.5 and GPT-4 to automate the data annotation process with automatic prompting techniques. The main aim of this paper is to reuse human annotation guidelines along with some annotated data to design automatic prompts for LLMs, focusing on the semantic proximity annotation task. Automatic prompts are compared to customized prompts. We further implement the prompting strategies into an open-source text annotation tool, enabling easy online use via the OpenAI API. Our study reveals the crucial role of accurate prompt design and suggests that prompting GPT-4 with human-like instructions is not straightforwardly possible for the semantic proximity task. We show that small modifications to the human guidelines already improve the performance, suggesting possible ways for future research.

cs.CL↗

arXiv2024

Emergence of quantum Griffiths singularity in disordered TiN thin films

The association of quantum Griffiths singularity (QGS) to the magnetic-field-induced superconductor-metal transition predicts the unconventional diverging behaviour of dynamical critical exponent in low disorder crystalline two-dimensional superconductors. But whether this state exists in the superconducting systems exhibiting superconductor-insulator transition remains elusive. Here, we report the emergence of quantum Griffiths singularity in ultrathin disordered TiN thin films with more than two orders of magnitude variation in their normal state resistance. For both superconductor-metal transition and superconductor-insulator transition types, a diverging critical exponent is observed while approaching the quantum phase transition. Further, the magnetoresistance isotherms obey a direct activated scaling governed by an infinite-randomness fixed critical point. Finally, this work establishes the robustness of the QGS phenomenon towards a wide range of temperature and also towards a wide range of disorder strength as correlated with the normal state resistance.

cond-mat.supr-con↗

arXiv2023

Probing electron-electron interaction along with superconducting fluctuations in disordered TiN thin films

Here, we demonstrate an interplay between superconducting fluctuations and electron-electron interaction (EEI) by low temperature magnetotransport measurements for a set of 2D disordered TiN thin films. While cooling down the sample, a characteristic temperature T* is obtained from the R(T) at which superconducting fluctuations start to appear. The upturn in R(T) above T* corresponds to weak localization (WL) and/or EEI. By the temperature and field dependences of the observed resistance, we show that the upturn in R(T) originates mainly from EEI with a negligible contribution from WL. Further, we have used the modified Larkins electron-electron attraction strength beta(T/Tc), containing a field induced pair breaking parameter, in the Maki-Thompson (MT) superconducting fluctuation term. Here, the temperature dependence of the beta(T/Tc) obtained from the magnetoresistance analysis shows a diverging behavior close to Tc and it remains almost constant at higher temperature within the limit of ln(T/Tc) < 1. Interestingly, the variation of beta(T/Tc) on the reduced temperature (T/Tc) offers a common trend which has been closely followed by all the concerned samples presented in this study. Finally, the temperature dependence of inverse phase scattering time , as obtained from the magnetoresistance analysis, clearly shows two different regimes; the first one close to Tc follows the Ginzburg-Landau relaxation rate , whereas, the second one at high temperature varies almost linearly with temperature indicating the dominance of inelastic electron-electron scattering for the dephasing mechanism. These two regimes are followed in a generic way by all the samples in spite of being grown under different growth conditions.

cond-mat.supr-con↗

arXiv2022

Deep Gaussian Processes for Air Quality Inference

Air pollution kills around 7 million people annually, and approximately 2.4 billion people are exposed to hazardous air pollution. Accurate, fine-grained air quality (AQ) monitoring is essential to control and reduce pollution. However, AQ station deployment is sparse, and thus air quality inference for unmonitored locations is crucial. Conventional interpolation methods fail to learn the complex AQ phenomena. This work demonstrates that Deep Gaussian Process models (DGPs) are a promising model for the task of AQ inference. We implement Doubly Stochastic Variational Inference, a DGP algorithm, and show that it performs comparably to the state-of-the-art models.

cs.LG↗

arXiv2021

Accessing Phase Slip Events in Nb Meander Wires

We report transport studies through Nb-based superconducting meander wires fabricated by focused ion beam (FIB) milling technique. The effect of meandering on quantum transport has been probed experimentally by a direct comparison with the pristine thin-film device before meandering. The normal metal (NM) to superconductor (SC) phase transition becomes a wide and multi-step transition by meandering. Below the transition temperature (Tc), the resistance-versus-temperature measurements reveal resistive tailing which is explained by the thermally activated phase slip (TAPS) mechanism. The TAPS fit indicates a selective region of the meander to be responsible for the resistive tailing. Besides, the phase slip (PS) mechanism in the meander is evident in its current-voltage characteristics that feature the stair-case type intermediate resistive steps during the SC-NM transition. The modulation of the intermediate resistive steps is investigated with respect to temperature and external magnetic field. It is observed that the PS events are facilitated by magnetic fields up to about 250 mT. Further, the critical current varies strongly on the temperature and magnetic field for T less than 0.5Tc and H less tahn 100 mT where it fluctuates in an oscillatory manner. Finally, Nb based meander structures can be promising candidates for future PS based studies and applications.

cond-mat.supr-con↗

arXiv2021

A Robust nitridation technique for fabrication of disordered superconducting TiN thin films featuring phase slip events

Disorder induced phase slip (PS) events appearing in the current voltage characteristics (IVCs) are reported for two-dimensional TiN thin films produced by a robust substrate mediated nitridation technique. Here, high temperature annealing of Ti/Si3N4 based metal/substrate assembly is the key to produce majority phase TiN accompanied by TiSi2 and elemental Si as minority phases. The method itself introduces different level of disorder intrinsically by tuning the amount of the non-superconducting minority phases that are controlled by annealing temperature (Ta) and the film thickness. The superconducting critical temperature (Tc) strongly depends on Ta and the maximum Tc obtained from the demonstrated technique is about 4.8 K for the thickness range of about 12 nm and above. Besides, the dynamics of IVCs get modulated by the appearance of intermediated resistive steps for decreased Ta and the steps get more prominent for reduced thickness. Further, the deviation in the temperature dependent critical current (Ic) from the Ginzburg-Landau theoretical limit varies strongly with the thickness. Finally, the Tc, intermediate resistive steps in the IVCs and the depairing current are observed to alter in a similar fashion with Ta and the thickness indicating the robustness of the synthesis process to fabricate disordered nitride-based superconductor.

cond-mat.supr-con↗

arXiv2020

Interface study of thermally driven chemical kinetics involved in Ti/Si3N4 based metal-substrate assembly by X-ray photoelectron spectroscopy

Diffusion mediated interaction in metal-substrate assembly during high temperature annealing leads to possible formation of new composite materials. Here, sputtered grown Ti films on Si3N4/Si substrate has been reported to produce titanium nitride and silicide based binary composites while undergoing high vacuum annealing process at temperatures 650°C and above. Diffusion of thermally decomposed Si and N atoms from Si3N4 and their subsequent chemical reaction with Ti have been probed by X-ray photo electron spectroscopy. For annealing at 800°C and above, most of the Si atoms show preferences to stay in elemental form rather than developing silicide phase with Ti. Whereas at lower annealing temperature, silicide becomes the dominant phase for decomposed Si atoms. However, N atoms react promptly with Ti and form TiN which appears as the majority phase for each of the studied annealing temperature. Further, the nitride and silicide phases across the films have been compared quantitatively for various annealing temperature and the maximum silicide formation is observed for the sample annealed at 780°C. Finally, the thermally anchored metal-substrate interaction mechanism can be exploited to fabricate disordered superconducting TiN films where TiSi2 and Si can be used to tune the level of disorder by altering the annealing temperature.

cond-mat.mtrl-sci↗

arXiv2019

Substrate mediated nitridation of niobium into superconducting Nb2N thin films for phase slip study

Here we report a novel nitridation technique for transforming niobium into hexagonal Nb2N which appears to be superconducting below 1K. The nitridation is achieved by high temperature annealing of Nb films grown on Si3N4/Si (100) substrate under high vacuum. The structural characterization directs the formation of a majority Nb2N phase while the morphology shows granular nature of the films. The temperature dependent resistance measurements reveal a wide metal-to-superconductor transition featuring two distinct transition regions. The region close to the normal state varies strongly with the film thickness, whereas, the second region in the vicinity of the superconducting state remains almost unaltered but exhibiting resistive tailing. The current-voltage characteristics also display wide transition embedded with intermediate resistive states originated by phase slip lines. The transition width in current and the number of resistive steps depend on film thickness and they both increase with decrease in thickness. The broadening in transition width is explained by progressive establishment of superconductivity through proximity coupled superconducting nano-grains while finite size effects and quantum fluctuation may lead to the resistive tailing. Finally, by comparing with Nb control samples, we emphasize that Nb2N offers unconventional superconductivity with promises in the field of phase slip based device applications.

cond-mat.supr-con↗