arXiv subjects

Aman Gupta

Publications and source records attributed to Aman Gupta.

At least 19 recordsLinked to original sources

arXiv2026

Quantum Secure Time Transfer for Satellites

We experimentally demonstrate an entanglement-based Quantum-Secure Time Transfer (QSTT) system in an emulated low Earth orbit satellite-to-ground channel using a type-0 Sagnac-based entangled-photon source. Our new QSTT system delivers a finite Quantum Key Distribution (QKD) key rate of approximately $3~$bits$~\text{s}^{-1}$ with a $10^{-10}$ security parameter, providing a $30$-fold increase in the QKD key rate compared to the state-of-the-art entanglement-based QKD system delivered by the Micius satellite. Beyond this high key rate outcome, novel to our QSTT system is a GPS-free clock synchronization, optimized use of QKD, embedded post-quantum security, and obfuscation of the system configuration via use of a pre-shared key. Collectively, these enhancements deliver the most efficient and secure deployment of QSTT to date, and point the way forward to high-accuracy ultra-secure time transfer in space.

quant-ph↗

arXiv2026

SpanUQ: Span-Level Uncertainty Quantification for Large Language Model Generation

Uncertainty estimation is essential not only for the trustworthy deployment of large language models (LLMs) but also as a foundation for self-refinement in LLM generation. However, existing approaches operate at suboptimal granularities: token-level scores lack semantic coherence, while sequence-level scores fail to localize errors. We formalize Span-Level Uncertainty Estimation (SLUE), a new task that targets the natural granularity for uncertainty: semantically coherent text spans, each conveying a single assessable unit of meaning. To address this task, we introduce SPANUQ, a lightweight probe that distills the uncertainty knowledge from expensive multi-sample inference into a single forward pass over LLM hidden states. SPANUQ employs a DETR-style span decoder to simultaneously detect spans and estimate their uncertainty via a Mixture of Beta distribution, trained with a principled combination of Beta NLL regression and contrastive ranking objectives. We construct SPANUQ-BENCH, the first span-level uncertainty benchmark comprising 20K prompts, 293K annotated spans, and continuous soft labels derived from multi-sample claim verification. Experiments on five LLM backbones show that SPANUQ consistently achieves the best span-level uncertainty quality, outperforming the strongest probe baseline and all sampling-based methods while being 10-20x faster. Its DETR-based span detector attains 0.910 F1, surpassing the best heuristic by 39.4%, enabling precise error localization that sequence-level methods cannot provide. The framework generalizes across five LLMs spanning two model families.

cs.CL↗

arXiv2026

Building Customer Support AI Agents at 100M-User Scale: An Evaluation-Driven Framework

The rapid rise in LLM capabilities has made AI agents increasingly viable across a broad range of tasks. Among the most promising applications is building production-ready customer-facing agents, a challenge that demands coordinated excellence in evaluation methodology, context engineering, training, and online measurement. Yet these critical pillars are typically developed in isolation, creating blind spots that only surface after deployment. In this paper, we present a unified framework that bridges offline development with online impact for customer support AI agents at Nubank, a company with 100M+ users. Our approach integrates several key components: (1) structured context engineering tailored to customer support agents, (2) systematic human-in-the-loop prompt iteration, (3) rigorous LLM judge evaluation with measured inter-rater agreement and GEPA optimization for consistency, and (4) ideation-to-production validation. A central insight is that evaluation-pipeline quality directly determines iteration velocity. We present results from five production deployments spanning distinct domains: card delivery, debt management, credit-limit support, card management, and product explanation. These deployments deliver consistent customer-satisfaction gains while substantially accelerating iteration. In our card-delivery deployment, large-scale A/B testing yields a 37 percentage-point improvement in AI transactional Net Promoter Score and a 29 percentage-point gain in self-service rate over prior agent variants, alongside a strong correlation between offline simulation metrics and online outcomes, demonstrating that eval-driven development reliably predicts production impact. On most use cases, AI satisfaction reaches within a few percentage points of expert human agents.

cs.CL↗

arXiv2026

Constraints on High-Frequency Gravitational Waves from Graviton-Photon Conversion in the M87 Galaxy

High-frequency gravitational waves, particularly in the range $f \gtrsim 10^{10}~\mathrm{Hz}$, represent a compelling probe of physics beyond the Standard Model. Due to the absence of direct detection methods in this frequency regime, alternative strategies may be pursued. One promising approach involves the conversion of gravitons into photons in the presence of magnetic fields, a process known as the inverse Gertsenshtein effect. In this study, we explore such graviton-to-photon conversions occurring within the magnetic field environment of the M87 galaxy, utilizing realistic models for the galactic magnetic field and plasma density structure. We use the broadband electromagnetic spectrum of M87, ranging from millimeter to TeV gamma rays, to search for hidden contributions from graviton-photon conversions. In the well-constrained frequency range $10^{10}$-$10^{27}~\mathrm{Hz}$, the lack of excess emission allows us to place improved bounds on the gravitational wave strain amplitude $h_c$ or on spectral energy density $\Omega_{\mathrm{gw}} h^2$. We find that our results from M87 yield substantially stronger constraints compared to existing bounds derived from Milky Way magnetic field considerations, with improvements ranging from one to five orders of magnitude depending on the frequency band, thereby enhancing the prospects for probing high-frequency gravitational wave backgrounds through indirect electromagnetic signatures.

hep-ph↗

arXiv2026

TRIM: Hybrid Inference via Targeted Stepwise Routing in Multi-Step Reasoning Tasks

Multi-step reasoning tasks like mathematical problem solving are vulnerable to cascading failures, where a single incorrect step leads to complete solution breakdown. Current LLM routing methods assign entire queries to one model, treating all reasoning steps as equal. We propose TRIM (Targeted routing in multi-step reasoning tasks), which routes only critical steps$\unicode{x2013}$those likely to derail the solution$\unicode{x2013}$to larger models while letting smaller models handle routine continuations. Our key insight is that targeted step-level interventions can fundamentally transform inference efficiency by confining expensive calls to precisely those steps where stronger models prevent cascading errors. TRIM operates at the step-level: it uses process reward models to identify erroneous steps and makes routing decisions based on step-level uncertainty and budget constraints. We develop several routing strategies within TRIM, ranging from a simple threshold-based policy to more expressive policies that reason about long-horizon accuracy-cost trade-offs and uncertainty in step-level correctness estimates. On MATH-500, even the simplest thresholding strategy surpasses prior routing methods with 5x higher cost efficiency, while more advanced policies match the strong, expensive model's performance using 80% fewer expensive model tokens. On harder benchmarks such as AIME, TRIM achieves up to 6x higher cost efficiency. All methods generalize effectively across math reasoning tasks, demonstrating that step-level difficulty represents fundamental characteristics of reasoning.

cs.AI↗

arXiv2025

Combined Quantum and Post-Quantum Security Performance Under Finite Keys

Recent advances in quantum-secure communication have highlighted the value of hybrid schemes that combine Quantum Key Distribution (QKD) with Post-Quantum Cryptography (PQC). Yet most existing hybrid designs omit realistic finite-key effects on QKD key rates and do not specify how to maintain security when both QKD and PQC primitives leak information through side-channels. These gaps limit the applicability of hybrid systems in practical, deployed networks. In this work, we advance a recently proposed hybrid QKD-PQC system by integrating tight finite-key security to the QKD primitive and improving the design for better scalability. This hybrid system employs an information-theoretically secure instruction sequence that determines the configurations of different primitives and thus ensures message confidentiality even when both the QKD and the PQC primitives are compromised. The novelty in our work lies in the implementation of the tightest finite-key security to date for the BBM92 protocol and the design improvements in the primitives of the hybrid system that ensure the processing time scales linearly with the size of secret instructions.

quant-ph↗

arXiv2025

Enhancing the sensitivity to neutrino oscillation parameters using synergy between T2K, NO$\nu$A and JUNO

We study the impact of combining the present NO$\nu$A and T2K data with simulated data from the JUNO experiment on the determination of the leptonic CP phase and the neutrino mass hierarchy. The current NO$\nu$A data exhibit a hierarchy--$\delta_{\rm CP}$ degeneracy, admitting both normal hierarchy (NH) with $\delta_{\rm CP} \in [0,180^\circ]$, and inverted hierarchy (IH) with $\delta_{\rm CP} \in [180^\circ,360^\circ]$ solutions at comparable significance, while T2K prefers $\delta_{\rm CP}\simeq 270^\circ$ for both hierarchies, leading to a $2\sigma$ tension between the two experiments for normal hierarchy. Using detailed GLoBES simulations, we show that future JUNO data with excellent hierarchy sensitivity, can lift the hierarchy--$\delta_{\rm CP}$ degeneracy in NO$\nu$A and strengthen the hierarchy reach of T2K in spite of having no $\delta_{\rm CP}$ sensitivity. Allowing the hierarchy to be a free parameter in the fit, if the true ordering is IH, JUNO aligns the NO$\nu$A and T2K allowed regions and resolves their present tension; if NH is true, the tension continues to persist. We also show that JUNO's precise measurement of $|\Delta_{31}|$ leads to improved constraints on $\sin^2\theta_{23}$ and $\delta_{\rm CP}$ for normal mass hierarchy in NO$\nu$A even though JUNO itself is insensitive to these parameters. Finally, updated solar parameter measurements from JUNO's first data release further enhance the combined precision. Our results demonstrate that JUNO plays a crucial synergistic role in the global neutrino oscillation programme, enabling a more robust determination of the mass ordering and improving the sensitivity to the CP phase when combined with long-baseline data.

hep-ph↗

arXiv2025

A New Quantum Secure Time Transfer System

High-precision clock synchronization is essential for a wide range of network-distributed applications. In the quantum space, these applications include communication, sensing, and positioning. However, current synchronization techniques are vulnerable to attacks, such as intercept-resend attacks, spoofing, and delay attacks. Here, we propose and experimentally demonstrate a new quantum secure time transfer (QSTT) system, subsequently used for clock synchronization, that largely negates such attacks. Novel to our system is the optimal use of self-generated quantum keys within the QSTT to information-theoretically secure the maximum amount of timing data; as well as the introduction, within a hybrid quantum/post-quantum architecture, of an information-theoretic secure obfuscated encryption sequence of the remaining timing data. With these enhancements, we argue that our new system represents the most robust implementation of QSTT to date.

quant-ph↗

arXiv2025

LANTERN: Scalable Distillation of Large Language Models for Job-Person Fit and Explanation

Large language models (LLMs) have achieved strong performance across a wide range of natural language processing tasks. However, deploying LLMs at scale for domain specific applications, such as job-person fit and explanation in job seeking platforms, introduces distinct challenges. At LinkedIn, the job person fit task requires analyzing a candidate's public profile against job requirements to produce both a fit assessment and a detailed explanation. Directly applying open source or finetuned LLMs to this task often fails to yield high quality, actionable feedback due to the complexity of the domain and the need for structured outputs. Moreover, the large size of these models leads to high inference latency and limits scalability, making them unsuitable for online use. To address these challenges, we introduce LANTERN, a novel LLM knowledge distillation framework tailored specifically for job person fit tasks. LANTERN involves modeling over multiple objectives, an encoder model for classification purpose, and a decoder model for explanation purpose. To better distill the knowledge from a strong black box teacher model to multiple downstream models, LANTERN incorporates multi level knowledge distillation that integrates both data and logit level insights. In addition to introducing the knowledge distillation framework, we share our insights on post training techniques and prompt engineering, both of which are crucial for successfully adapting LLMs to domain specific downstream tasks. Extensive experimental results demonstrate that LANTERN significantly improves task specific metrics for both job person fit and explanation. Online evaluations further confirm its effectiveness, showing measurable gains in job seeker engagement, including a 0.24\% increase in apply rate and a 0.28\% increase in qualified applications.

cs.CL↗

arXiv2025

VAL-Bench: Belief Consistency as a measure for Value Alignment in Language Models

Large language models (LLMs) are increasingly being used for tasks where outputs shape human decisions, so it is critical to verify that their responses consistently reflect desired human values. Humans, as individuals or groups, don't agree on a universal set of values, which makes evaluating value alignment difficult. Existing benchmarks often use hypothetical or commonsensical situations, which don't capture the complexity and ambiguity of real-life debates. We introduce the Value ALignment Benchmark (VAL-Bench), which measures the consistency in language model belief expressions in response to real-life value-laden prompts. VAL-Bench consists of 115K pairs of prompts designed to elicit opposing stances on a controversial issue, extracted from Wikipedia. We use an LLM-as-a-judge, validated against human annotations, to evaluate if the pair of responses consistently expresses either a neutral or a specific stance on the issue. Applied across leading open- and closed-source models, the benchmark shows considerable variation in consistency rates (ranging from ~10% to ~80%), with Claude models the only ones to achieve high levels of consistency. Lack of consistency in this manner risks epistemic harm by making user beliefs dependent on how questions are framed rather than on underlying evidence, and undermines LLM reliability in trust-critical applications. Therefore, we stress the importance of research towards training belief consistency in modern LLMs. By providing a scalable, reproducible benchmark, VAL-Bench enables systematic measurement of necessary conditions for value alignment.

cs.AI↗

arXiv2025

Effective Quantization of Muon Optimizer States

The Muon optimizer, based on matrix orthogonalization, has recently shown faster convergence and better computational efficiency over AdamW in LLM pre-training. However, the memory overhead of maintaining high-precision optimizer states remains a challenge for large-scale deployment. In this paper, we introduce the 8-bit Muon optimizer using blockwise quantization. In extensive Chinchilla-optimal experiments on pre-training models of up to 2.7B in size and fine-tuning them for instruction following, we demonstrate that 8-bit Muon achieves parity with Muon in terms of validation loss and downstream benchmarks, while achieving up to a 62\% reduction in optimizer state footprint. Crucially, we show that Muon's update mechanism is uniquely compatible with a simple linear quantization scheme, bypassing the complex dynamic scaling required for quantized AdamW. We supplement our empirical findings with a theoretical analysis of Muon's robustness to quantization noise.

cs.LG↗

arXiv2025

Finetuning AI Foundation Models to Develop Subgrid-Scale Parameterizations: A Case Study on Atmospheric Gravity Waves

Global climate models parameterize a range of atmospheric-oceanic processes like gravity waves, clouds, moist convection, and turbulence that cannot be sufficiently resolved. These subgrid-scale closures for unresolved processes are a leading source of model uncertainty. Here, we present a new approach to developing machine learning parameterizations of small-scale climate processes by fine-tuning a pre-trained AI foundation model (FM). FMs are largely unexplored in climate research. A pre-trained encoder-decoder from a 2.3 billion parameter FM (NASA and IBM Research's Prithvi WxC) -- which contains a latent probabilistic representation of atmospheric evolution -- is fine-tuned (or reused) to create a deep learning parameterization for atmospheric gravity waves (GWs). The parameterization captures GW effects for a coarse-resolution climate model by learning the fluxes from an atmospheric reanalysis with 10 times finer resolution. A comparison of monthly averages and instantaneous evolution with a machine learning model baseline (an Attention U-Net) reveals superior predictive performance of the FM parameterization throughout the atmosphere, even in regions excluded from pre-training. This performance boost is quantified using the Hellinger distance, which is 0.11 for the baseline and 0.06 for the fine-tuned model. Our findings emphasize the versatility and reusability of FMs, which could be used to accomplish a range of atmosphere- and climate-related applications, leading the way for the creation of observations-driven and physically accurate parameterizations for more earth-system processes.

physics.ao-ph↗

arXiv2025

MAUSAM: An Observations-focused assessment of Global AI Weather Prediction Models During the South Asian Monsoon

Accurate weather forecasts are critical for societal planning and disaster preparedness. Yet these forecasts remain challenging to produce and evaluate, especially in regions with sparse observational coverage. Current evaluation of artificial intelligence (AI) weather prediction relies primarily on reanalyses, which can obscure important deficiencies. Here we present MAUSAM (Measuring AI Uncertainty during South Asian Monsoon), an evaluation of seven leading AI-based forecasting systems - FourCastNet, FourCastNet-SFNO, Pangu-Weather, GraphCast, Aurora, AIFS, and GenCast - during the South Asian Monsoon, using ground-based weather stations, rain gauge networks, and geostationary satellite imagery. The AI models demonstrate impressive forecast skill during monsoon across a broad range of variables, ranging from large-scale surface temperature and winds to precipitation, cloud cover, and subseasonal to seasonal eddy statistics, highlighting the strength of data-driven weather prediction. However, the models still exhibit systematic errors at finer scales like the underprediction of extreme precipitation, divergent cyclone tracks, and the mesoscale kinetic energy spectra, highlighting avenues for future improvement. A comparison against observations reveals forecast errors up to 15-45% larger than those relative to reanalysis and traditional forecasts, indicating that reanalysis-centric benchmarks can overstate forecast skill. Of the models assessed, AIFS achieves the most consistent representation of atmospheric variables, with GraphCast and GenCast also showing strong skill. The analysis presents a framework for evaluating AI weather models on regional prediction and highlights both the promise and current limitations of AI weather prediction in data-sparse regions, underscoring the importance of observational evaluation for future operational adoption.

physics.ao-ph↗

arXiv2025

Obfuscated Quantum and Post-Quantum Cryptography

In this work, we present an experimental deployment of a new design for combined quantum key distribution (QKD) and post-quantum cryptography (PQC). Novel to our system is the dynamic obfuscation of the QKD-PQC sequence of operations, the number of operations, and parameters related to the operations; coupled to the integration of a GPS-free quantum synchronization protocol within the QKD process. We compare the performance and overhead of our QKD-PQC system relative to a standard QKD system with one-time pad encryption, demonstrating that our design can operate in real time with little additional overhead caused by the new security features. Since our system can offer additional defensive strategies against a wide spectrum of practical attacks that undermine deployed QKD, PQC, and certain combinations of these two primitives, we suggest that our design represents one of the most secure communication systems currently available. Given the dynamic nature of its obfuscation attributes, our new system can also be adapted in the field to defeat yet-to-be-discovered practical attacks.

quant-ph↗

arXiv2025

Your Spending Needs Attention: Modeling Financial Habits with Transformers

Predictive models play a crucial role in the financial industry, enabling risk prediction, fraud detection, and personalized recommendations, where slight changes in core model performance can result in billions of dollars in revenue or losses. While financial institutions have access to enormous amounts of user data (e.g., bank transactions, in-app events, and customer support logs), leveraging this data effectively remains challenging due to its complexity and scale. Thus, in many financial institutions, most production models follow traditional machine learning (ML) approaches by converting unstructured data into manually engineered tabular features. Conversely, other domains (e.g., natural language processing) have effectively utilized self-supervised learning (SSL) to learn rich representations from raw data, removing the need for manual feature extraction. In this paper, we investigate using transformer-based representation learning models for transaction data, hypothesizing that these models, trained on massive data, can provide a novel and powerful approach to understanding customer behavior. We propose a new method enabling the use of SSL with transaction data by adapting transformer-based models to handle both textual and structured attributes. Our approach, denoted nuFormer, includes an end-to-end fine-tuning method that integrates user embeddings with existing tabular features. Our experiments demonstrate improvements for large-scale recommendation problems at Nubank. Notably, these gains are achieved solely through enhanced representation learning rather than incorporating new data sources.

cs.IR↗

arXiv2025

How and Where to Translate? The Impact of Translation Strategies in Cross-lingual LLM Prompting

Despite advances in the multilingual capabilities of Large Language Models (LLMs), their performance varies substantially across different languages and tasks. In multilingual retrieval-augmented generation (RAG)-based systems, knowledge bases (KB) are often shared from high-resource languages (such as English) to low-resource ones, resulting in retrieved information from the KB being in a different language than the rest of the context. In such scenarios, two common practices are pre-translation to create a mono-lingual prompt and cross-lingual prompting for direct inference. However, the impact of these choices remains unclear. In this paper, we systematically evaluate the impact of different prompt translation strategies for classification tasks with RAG-enhanced LLMs in multilingual systems. Experimental results show that an optimized prompting strategy can significantly improve knowledge sharing across languages, therefore improve the performance on the downstream classification task. The findings advocate for a broader utilization of multilingual resource sharing and cross-lingual prompt optimization for non-English languages, especially the low-resource ones.

cs.CL↗

arXiv2025

Multilingual Information Retrieval with a Monolingual Knowledge Base

Multilingual information retrieval has emerged as powerful tools for expanding knowledge sharing across languages. On the other hand, resources on high quality knowledge base are often scarce and in limited languages, therefore an effective embedding model to transform sentences from different languages into a feature vector space same as the knowledge base language becomes the key ingredient for cross language knowledge sharing, especially to transfer knowledge available in high-resource languages to low-resource ones. In this paper we propose a novel strategy to fine-tune multilingual embedding models with weighted sampling for contrastive learning, enabling multilingual information retrieval with a monolingual knowledge base. We demonstrate that the weighted sampling strategy produces performance gains compared to standard ones by up to 31.03\% in MRR and up to 33.98\% in Recall@3. Additionally, our proposed methodology is language agnostic and applicable for both multilingual and code switching use cases.

cs.CL↗

arXiv2025

On the Robustness of Reward Models for Language Model Alignment

The Bradley-Terry (BT) model is widely practiced in reward modeling for reinforcement learning with human feedback (RLHF). Despite its effectiveness, reward models (RMs) trained with BT model loss are prone to over-optimization, losing generalizability to unseen input distributions. In this paper, we study the cause of over-optimization in RM training and its downstream effects on the RLHF procedure, accentuating the importance of distributional robustness of RMs in unseen data. First, we show that the excessive dispersion of hidden state norms is the main source of over-optimization. Then, we propose batch-wise sum-to-zero regularization (BSR) to enforce zero-centered reward sum per batch, constraining the rewards with extreme magnitudes. We assess the impact of BSR in improving robustness in RMs through four scenarios of over-optimization, where BSR consistently manifests better robustness. Subsequently, we compare the plain BT model and BSR on RLHF training and empirically show that robust RMs better align the policy to the gold preference model. Finally, we apply BSR to high-quality data and models, which surpasses state-of-the-art RMs in the 8B scale by adding more than 5% in complex preference prediction tasks. By conducting RLOO training with 8B RMs, AlpacaEval 2.0 reduces generation length by 40% while adding a 7% increase in win rate, further highlighting that robustness in RMs induces robustness in RLHF training. We release the code, data, and models: https://github.com/LinkedIn-XFACT/RM-Robustness.

cs.CL↗