arXiv subjects

Saptarshi Sengupta

Publications and source records attributed to Saptarshi Sengupta.

At least 19 recordsLinked to original sources

arXiv2026

LLM Unlearning for Cyber Defense: A Survey on Methods, Challenges, and Emerging Threats

LLMs are increasingly deployed in security-critical systems across healthcare, finance, education, and decision support, yet their inability to forget creates serious cybersecurity, privacy, and safety risks. Sensitive personal information, copyrighted material, hazardous domain knowledge, and memorized training data remain encoded across billions of parameters long after deployment, leaving models vulnerable to extraction, jailbreak attacks, membership inference, and regulatory non-compliance. Real-world incidents, from chatbots regenerating private information to fabricated legal citations producing direct legal and financial cost, place the problem at the center of the emerging-threats landscape rather than the realm of speculation. Because retraining billion-parameter models on revised corpora is computationally infeasible, and because knowledge within an LLM is distributed and entangled across parameters rather than localized to identifiable units, LLM unlearning has emerged as the principal cyber defense response, aiming to remove or suppress targeted knowledge from a trained model without retraining and without eroding what the model should still know. A central question, however, remains unresolved. Do current methods genuinely remove knowledge, or do they only stop the model from expressing it under ordinary prompting conditions? This survey examines LLM unlearning through the lens of security, robustness, and verifiable forgetting, with primary focus on gradient-based methods, which have come to dominate the field due to their compatibility with existing training pipelines and their scalability to billion-parameter models.

cs.LG↗

arXiv2026

MorphStrata: Layer-Specific Perturbations for Generating Morphence Students in Time-Series Moving Target Defense

Time-series forecasting models remain vulnerable to gradient-based adversarial attacks while existing defense mechanisms typically incur a trade-off in robustness for bounded response and compute cost. The problem is pronounced in Moving Target Defense where maintaining multiple randomized model instances substantially exacerbates the training overhead. In this work, we introduce MorphStrata, a student generation strategy with selective, layer-specific stochastic noise injection that extends the traditional Morphence defense. MorphStrata uses a Transformer backbone as the teacher and perturbs randomly selected architectural blocks to create structured heterogeneity across student models in response to varied data distributions and threat models. We evaluate against vanilla Transformer and Morphence backbones on a suite of benchmarks including the Jena Climate, Electricity Load Diagrams, and Appliances Energy Prediction using FGSM, BIM and PGD attacks across multiple attack strengths. Across datasets and attack regimes, the proposed ensemble maintains comparable adversarial RMSE. Specifically, for high entropy, periodic datasets as in the case of the AEP data, MorphStrata achieves the lowest RMSE across all attacks and perturbation budgets, improving over the static baseline by up to 24.11% and 97.97% under FGSM and BIM respectively at an epsilon value of 0.5 over 30 randomized trials. Targeting the layers to generate MorphStrata students accounts for less than 1% increase in train-times over the Morphence MTD baseline for most of the experiments, while accounting for double digit gains in adversarial RMSE reduction. We also observe a positive correlation between higher pairwise L2 distance (among generated students) and overall defense effectiveness. In summary, MorphStrata maintains adversarial robustness as an MTD defense at marginal cost deltas when compared to existing baselines.

cs.LG↗

arXiv2026

Can Hallucinations Be Useful? Solving Multi-Hop Questions With SLMs By Chaining System-I/II Reasoning

Recently, there has been increased interest in Small Language Models (SLMs), which are fast, show good performance, and have lower hardware demands than large language models (LLMs). However, SLMs hallucinate more frequently than LLMs, impacting their ability to solve complex multi-step reasoning problems as early mistakes cascade to the final response. To address this, existing works think-first followed by iterative retrieval to reduce hallucination. We argue that the think-first strategy is not always necessary as we find that: (i) SLMs are often accurately confident in their initial answer and, (ii) hallucinations can actually be beneficial for honing in on the true answer. As such, we position our work as an inversion of this strategy, i.e., answer first-reason later. We propose a cognitively-inspired framework where the model is first allowed to quickly answer the question (System-I (zero-shot)) and then resorts to deeper thinking (System-II) based on evidence retrieved from a knowledge source using the initial hypothesis. By combining System-I and System-II style thinking, we show that our method can outperform prior work that takes the traditional think-first route on various multi-step question-answering benchmarks.

cs.CL↗

arXiv2026

GaLoRA: Parameter-Efficient Graph-Aware LLMs for Node Classification

The rapid rise of large language models (LLMs) and their ability to capture semantic relationships has led to their adoption in a wide range of applications. Text-attributed graphs (TAGs) are a notable example where LLMs can be combined with Graph Neural Networks to improve the performance of node classification. In TAGs, each node is associated with textual content and such graphs are commonly seen in various domains such as social networks, citation graphs, recommendation systems, etc. Effectively learning from TAGs would enable better representations of both structural and textual representations of the graph and improve decision-making in relevant domains. We present GaLoRA, a parameter-efficient framework that integrates structural information into LLMs. GaLoRA demonstrates competitive performance on node classification tasks with TAGs, performing on par with state-of-the-art models with just 0.24% of the parameter count required by full LLM fine-tuning. We experiment with three real-world datasets to showcase GaLoRA's effectiveness in combining structural and semantical information on TAGs.

cs.LG↗

arXiv2025

ToolDreamer: Instilling LLM Reasoning Into Tool Retrievers

Tool calling has become increasingly popular for Large Language Models (LLMs). However, for large tool sets, the resulting tokens would exceed the LLM's context window limit, making it impossible to include every tool. Hence, an external retriever is used to provide LLMs with the most relevant tools for a query. Existing retrieval models rank tools based on the similarity between a user query and a tool description (TD). This leads to suboptimal retrieval as user requests are often poorly aligned with the language of TD. To remedy the issue, we propose ToolDreamer, a framework to condition retriever models to fetch tools based on hypothetical (synthetic) TD generated using an LLM, i.e., description of tools that the LLM feels will be potentially useful for the query. The framework enables a more natural alignment between queries and tools within the language space of TD's. We apply ToolDreamer on the ToolRet dataset and show that our method improves the performance of sparse and dense retrievers with and without training, thus showcasing its flexibility. Through our proposed framework, our aim is to offload a portion of the reasoning burden to the retriever so that the LLM may effectively handle a large collection of tools without inundating its context window.

cs.CL↗

arXiv2025

Adversarial Attacks and Defenses in Multivariate Time-Series Forecasting for Smart and Connected Infrastructures

The emergence of deep learning models has revolutionized various industries over the last decade, leading to a surge in connected devices and infrastructures. However, these models can be tricked into making incorrect predictions with high confidence, leading to disastrous failures and security concerns. To this end, we explore the impact of adversarial attacks on multivariate time-series forecasting and investigate methods to counter them. Specifically, we employ untargeted white-box attacks, namely the Fast Gradient Sign Method (FGSM) and the Basic Iterative Method (BIM), to poison the inputs to the training process, effectively misleading the model. We also illustrate the subtle modifications to the inputs after the attack, which makes detecting the attack using the naked eye quite difficult. Having demonstrated the feasibility of these attacks, we develop robust models through adversarial training and model hardening. We are among the first to showcase the transferability of these attacks and defenses by extrapolating our work from the benchmark electricity data to a larger, 10-year real-world data used for predicting the time-to-failure of hard disks. Our experimental results confirm that the attacks and defenses achieve the desired security thresholds, leading to a 72.41% and 94.81% decrease in RMSE for the electricity and hard disk datasets respectively after implementing the adversarial defenses.

cs.LG↗

arXiv2025

BioMol-MQA: A Multi-Modal Question Answering Dataset For LLM Reasoning Over Bio-Molecular Interactions

Retrieval augmented generation (RAG) has shown great power in improving Large Language Models (LLMs). However, most existing RAG-based LLMs are dedicated to retrieving single modality information, mainly text; while for many real-world problems, such as healthcare, information relevant to queries can manifest in various modalities such as knowledge graph, text (clinical notes), and complex molecular structure. Thus, being able to retrieve relevant multi-modality domain-specific information, and reason and synthesize diverse knowledge to generate an accurate response is important. To address the gap, we present BioMol-MQA, a new question-answering (QA) dataset on polypharmacy, which is composed of two parts (i) a multimodal knowledge graph (KG) with text and molecular structure for information retrieval; and (ii) challenging questions that designed to test LLM capabilities in retrieving and reasoning over multimodal KG to answer questions. Our benchmarks indicate that existing LLMs struggle to answer these questions and do well only when given the necessary background data, signaling the necessity for strong RAG frameworks.

cs.CL↗

arXiv2025

Advancing Decoding Strategies: Enhancements in Locally Typical Sampling for LLMs

This chapter explores advancements in decoding strategies for large language models (LLMs), focusing on enhancing the Locally Typical Sampling (LTS) algorithm. Traditional decoding methods, such as top-k and nucleus sampling, often struggle to balance fluency, diversity, and coherence in text generation. To address these challenges, Adaptive Semantic-Aware Typicality Sampling (ASTS) is proposed as an improved version of LTS, incorporating dynamic entropy thresholding, multi-objective scoring, and reward-penalty adjustments. ASTS ensures contextually coherent and diverse text generation while maintaining computational efficiency. Its performance is evaluated across multiple benchmarks, including story generation and abstractive summarization, using metrics such as perplexity, MAUVE, and diversity scores. Experimental results demonstrate that ASTS outperforms existing sampling techniques by reducing repetition, enhancing semantic alignment, and improving fluency.

cs.CL↗

arXiv2025

MAG-V: A Multi-Agent Framework for Synthetic Data Generation and Verification

Extending the capabilities of Large Language Models (LLMs) with functions or tools for environment interaction has led to the emergence of the agent paradigm. In industry, training an LLM is not always feasible because of the scarcity of domain data, legal holds on proprietary customer data, rapidly changing business requirements, and the need to prototype new assistants. Agents provide an elegant solution to the above by relying on the zero-shot reasoning abilities of the underlying LLM and utilizing tools to explore and reason over customer data and respond to user requests. However, there are two concerns here: (I) acquiring large scale customer queries for agent testing is time-consuming, and (II) high reliance on the tool call sequence (or trajectory) followed by the agent to respond to user queries may lead to unexpected or incorrect behavior. To address this, we propose MAG-V, a multi-agent framework to first generate a dataset of questions that mimic customer queries; and second, reverse-engineer alternate questions from the responses for trajectory verification. Initial results indicate that our synthetic data can improve agent performance on actual customer queries. Furthermore, our trajectory verification methodology, inspired by distant supervision and using traditional machine learning (ML) models, outperforms a GPT-4o judge baseline by 11% accuracy and matches the performance of a GPT-4 judge on our constructed dataset. Overall, our approach is a step towards unifying diverse task agents into a cohesive framework for achieving an aligned objective.

cs.CL↗

arXiv2024

Towards Efficient Methods in Medical Question Answering using Knowledge Graph Embeddings

In Natural Language Processing (NLP), Machine Reading Comprehension (MRC) is the task of answering a question based on a given context. To handle questions in the medical domain, modern language models such as BioBERT, SciBERT and even ChatGPT are trained on vast amounts of in-domain medical corpora. However, in-domain pre-training is expensive in terms of time and resources. In this paper, we propose a resource-efficient approach for injecting domain knowledge into a model without relying on such domain-specific pre-training. Knowledge graphs are powerful resources for accessing medical information. Building on existing work, we introduce a method using Multi-Layer Perceptrons (MLPs) for aligning and integrating embeddings extracted from medical knowledge graphs with the embedding spaces of pre-trained language models (LMs). The aligned embeddings are fused with open-domain LMs BERT and RoBERTa that are fine-tuned for two MRC tasks, span detection (COVID-QA) and multiple-choice questions (PubMedQA). We compare our method to prior techniques that rely on a vocabulary overlap for embedding alignment and show how our method circumvents this requirement to deliver better performance. On both datasets, our method allows BERT/RoBERTa to either perform on par (occasionally exceeding) with stronger domain-specific models or show improvements in general over prior techniques. With the proposed approach, we signal an alternative method to in-domain pre-training to achieve domain proficiency. Our code is available here.

cs.CL↗

arXiv2024

TOP-Training: Target-Oriented Pretraining for Medical Extractive Question Answering

We study extractive question-answering in the medical domain (Medical-EQA). This problem has two main challenges: (i) domain specificity, as most AI models lack necessary domain knowledge, and (ii) extraction-based answering style, which restricts most autoregressive LLMs due to potential hallucinations. To handle those challenges, we propose TOP-Training, a target-oriented pre-training paradigm that stands out among all domain adaptation techniques with two desirable features: (i) TOP-Training moves one step further than popular domain-oriented fine-tuning since it not only moves closer to the target domain, but also familiarizes itself with the target dataset, and (ii) it does not assume the existence of a large set of unlabeled instances from the target domain. Specifically, for a target Medical-EQA dataset, we extract its entities and leverage large language models (LLMs) to generate synthetic texts containing those entities; we then demonstrate that pretraining on this synthetic text data yields better performance on the target Medical-EQA benchmarks. Overall, our contributions are threefold: (i) TOP-Training, a new pretraining technique to effectively adapt LLMs to better solve a target problem, (ii) TOP-Training has a wide application scope because it does not require the target problem to have a large set of unlabeled data, and (iii) our experiments highlight the limitations of autoregressive LLMs, emphasizing TOP-Training as a means to unlock the true potential of bidirectional LLMs.

cs.CL↗

arXiv2024

Exploring Language Model Generalization in Low-Resource Extractive QA

In this paper, we investigate Extractive Question Answering (EQA) with Large Language Models (LLMs) under domain drift, i.e., can LLMs generalize to domains that require specific knowledge such as medicine and law in a zero-shot fashion without additional in-domain training? To this end, we devise a series of experiments to explain the performance gap empirically. Our findings suggest that: (a) LLMs struggle with dataset demands of closed domains such as retrieving long answer spans; (b) Certain LLMs, despite showing strong overall performance, display weaknesses in meeting basic requirements as discriminating between domain-specific senses of words which we link to pre-processing decisions; (c) Scaling model parameters is not always effective for cross domain generalization; and (d) Closed-domain datasets are quantitatively much different than open-domain EQA datasets and current LLMs struggle to deal with them. Our findings point out important directions for improving existing LLMs.

cs.CL↗

arXiv2024

The State of Lithium-Ion Battery Health Prognostics in the CPS Era

Lithium-ion batteries (Li-ion) have revolutionized energy storage technology, becoming integral to our daily lives by powering a diverse range of devices and applications. Their high energy density, fast power response, recyclability, and mobility advantages have made them the preferred choice for numerous sectors. This paper explores the seamless integration of Prognostics and Health Management within batteries, presenting a multidisciplinary approach that enhances the reliability, safety, and performance of these powerhouses. Remaining useful life (RUL), a critical concept in prognostics, is examined in depth, emphasizing its role in predicting component failure before it occurs. The paper reviews various RUL prediction methods, from traditional models to cutting-edge data-driven techniques. Furthermore, it highlights the paradigm shift toward deep learning architectures within the field of Li-ion battery health prognostics, elucidating the pivotal role of deep learning in addressing battery system complexities. Practical applications of PHM across industries are also explored, offering readers insights into real-world implementations.This paper serves as a comprehensive guide, catering to both researchers and practitioners in the field of Li-ion battery PHM.

cs.LG↗

arXiv2024

Milestones in Bengali Sentiment Analysis leveraging Transformer-models: Fundamentals, Challenges and Future Directions

Sentiment Analysis (SA) refers to the task of associating a view polarity (usually, positive, negative, or neutral; or even fine-grained such as slightly angry, sad, etc.) to a given text, essentially breaking it down to a supervised (since we have the view labels apriori) classification task. Although heavily studied in resource-rich languages such as English thus pushing the SOTA by leaps and bounds, owing to the arrival of the Transformer architecture, the same cannot be said for resource-poor languages such as Bengali (BN). For a language spoken by roughly 300 million people, the technology enabling them to run trials on their favored tongue is severely lacking. In this paper, we analyze the SOTA for SA in Bengali, particularly, Transformer-based models. We discuss available datasets, their drawbacks, the nuances associated with Bengali i.e. what makes this a challenging language to apply SA on, and finally provide insights for future direction to mitigate the limitations in the field.

cs.CL↗

arXiv2023

De-SaTE: Denoising Self-attention Transformer Encoders for Li-ion Battery Health Prognostics

The usage of Lithium-ion (Li-ion) batteries has gained widespread popularity across various industries, from powering portable electronic devices to propelling electric vehicles and supporting energy storage systems. A central challenge in Li-ion battery reliability lies in accurately predicting their Remaining Useful Life (RUL), which is a critical measure for proactive maintenance and predictive analytics. This study presents a novel approach that harnesses the power of multiple denoising modules, each trained to address specific types of noise commonly encountered in battery data. Specifically, a denoising auto-encoder and a wavelet denoiser are used to generate encoded/decomposed representations, which are subsequently processed through dedicated self-attention transformer encoders. After extensive experimentation on NASA and CALCE data, a broad spectrum of health indicator values are estimated under a set of diverse noise patterns. The reported error metrics on these data are on par with or better than the state-of-the-art reported in recent literature.

cs.LG↗

arXiv2023

Investigating Antigram Behaviour using Distributional Semantics

The field of computational linguistics constantly presents new challenges and topics for research. Whether it be analyzing word usage changes over time or identifying relationships between pairs of seemingly unrelated words. To this point, we identify Anagrams and Antigrams as words possessing such unique properties. The presented work is an exploration into generating anagrams from a given word and determining whether there exists antigram (semantically opposite anagrams) relationships between the pairs of generated anagrams using GloVe embeddings. We propose a rudimentary, yet interpretable, rule-based algorithm for detecting antigrams. On a small dataset of just 12 antigrams, our approach yielded an accuracy of 39\% which shows that there is much work left to be done in this space.

cs.CL↗

arXiv2023

TFBEST: Dual-Aspect Transformer with Learnable Positional Encoding for Failure Prediction

Hard Disk Drive (HDD) failures in datacenters are costly - from catastrophic data loss to a question of goodwill, stakeholders want to avoid it like the plague. An important tool in proactively monitoring against HDD failure is timely estimation of the Remaining Useful Life (RUL). To this end, the Self-Monitoring, Analysis and Reporting Technology employed within HDDs (S.M.A.R.T.) provide critical logs for long-term maintenance of the security and dependability of these essential data storage devices. Data-driven predictive models in the past have used these S.M.A.R.T. logs and CNN/RNN based architectures heavily. However, they have suffered significantly in providing a confidence interval around the predicted RUL values as well as in processing very long sequences of logs. In addition, some of these approaches, such as those based on LSTMs, are inherently slow to train and have tedious feature engineering overheads. To overcome these challenges, in this work we propose a novel transformer architecture - a Temporal-fusion Bi-encoder Self-attention Transformer (TFBEST) for predicting failures in hard-drives. It is an encoder-decoder based deep learning technique that enhances the context gained from understanding health statistics sequences and predicts a sequence of the number of days remaining before a disk potentially fails. In this paper, we also provide a novel confidence margin statistic that can help manufacturers replace a hard-drive within a time frame. Experiments on Seagate HDD data show that our method significantly outperforms the state-of-the-art RUL prediction methods during testing over the exhaustive 10-year data from Backblaze (2013-present). Although validated on HDD failure prediction, the TFBEST architecture is well-suited for other prognostics applications and may be adapted for allied regression problems.

cs.LG↗

arXiv2023

Spatio-temporal Storytelling? Leveraging Generative Models for Semantic Trajectory Analysis

In this paper, we lay out a vision for analysing semantic trajectory traces and generating synthetic semantic trajectory data (SSTs) using generative language model. Leveraging the advancements in deep learning, as evident by progress in the field of natural language processing (NLP), computer vision, etc. we intend to create intelligent models that can study the semantic trajectories in various contexts, predicting future trends, increasing machine understanding of the movement of animals, humans, goods, etc. enhancing human-computer interactions, and contributing to an array of applications ranging from urban-planning to personalized recommendation engines and business strategy.

cs.CL↗