arXiv subjects

Sourav Saha

Publications and source records attributed to Sourav Saha.

At least 19 recordsLinked to original sources

arXiv2026

Right Knowledge, Wrong Answer: Characterizing Parametric Temporal Conflict in Open-Weight Language Models

Language models may encode both outdated facts and their newer replacements. We introduce Parametric Temporal Conflict (PTC), where the newer fact is present and recoverable, but the default forward pass prefers the outdated one. We release a deterministically verified benchmark of 8,746 Wikidata position-holder transitions and evaluate four open-weight language models across three families. A date-prefix prompt recovers the newer fact in 61-81% of PTC cases. Activation patching flips predictions in 72-85% of cases and localizes the preference to model-specific upper-layer regions. Residual-stream steering outperforms norm-matched random directions, indicating direction-specific representations. These results show that PTC reflects a localized representational preference rather than missing knowledge. Recovery is measured on oracle-identified conflicts because automatic detection remains unreliable. We release the benchmark, code, and statistics.

cs.LG↗

arXiv2026

Separable Neural Architectures as Physical World Models: from Mathematical Theory to Applications

This work introduces the Separable Neural Architecture (SNA), a function representational class combining neural approximation with tensor decomposition. The SNA decouples localized coordinate functions (atoms) from global interactions governed by a sparse, low-rank interaction object. This architecture possesses a compact and smooth inductive bias well-suited for solving partial differential equations (PDEs). When viewed as a Galerkin trial space under the variational SNA (VSNA) framework, the formulation satisfies classical variational guarantees under Lax-Milgram: well-posedness, quasi-optimality, convergence, and stability. In high-dimensional spatiotemporal--parametric PDEs, the VSNA mitigates the curse of dimensionality by scaling algebraically rather than exponentially. Exploiting an entirely factorized, tensor-native alternating least squares (ALS) optimization framework reduces this cost to linear in dimension. The VSNA is validated across elliptic, hyperbolic, and parabolic systems, demonstrating close alignment with predicted algebraic and spectral scaling rates. We showcase the SNA as a "solve once, query anywhere" physical world model via two engineering case studies: a 7D parametric manufacturing simulation and an experimental thermal-to-property inversion pipeline for Inconel 718. The VSNA executes a 1,000,000-query Monte Carlo sweep in 102s on a standard laptop CPU, yielding a 150,000x speedup over a full-grid finite element baseline hosted on an NVIDIA A100 GPU. It further enables real-time generative inverse-mode reconstructions under 100ms. These results demonstrate that the SNA serves as a compact mathematical substrate for continuous parameter manifolds to enable real-time inversion, optimization loops, and rapid uncertainty propagation.

cs.LG↗

arXiv2026

Separable neural architectures as a primitive for unified predictive and generative intelligence

Intelligent systems across physics, language and perception often exhibit factorisable structure, yet are typically modelled by monolithic neural architectures that do not explicitly exploit this structure. The separable neural architecture (SNA) addresses this by formalising a representational class that unifies additive, quadratic and tensor-decomposed neural models. By constraining interaction order and tensor rank, SNAs impose a structural inductive bias that factorises high-dimensional mappings into low-arity components. Separability need not be a property of the system itself: it often emerges in the coordinates or representations through which the system is expressed. Crucially, this coordinate-aware formulation reveals a structural analogy between chaotic spatiotemporal dynamics and linguistic autoregression. By treating continuous physical states as smooth, separable embeddings, SNAs enable distributional modelling of chaotic systems. This approach mitigates the nonphysical drift characteristics of deterministic operators whilst remaining applicable to discrete sequences. The compositional versatility of this approach is demonstrated across four domains: autonomous waypoint navigation via reinforcement learning, inverse generation of multifunctional microstructures, distributional modelling of turbulent flow and neural language modelling. These results establish the separable neural architecture as a domain-agnostic primitive for predictive and generative intelligence, capable of unifying both deterministic and distributional representations.

cs.LG↗

arXiv2026

LiCQA : A Lightweight Complex Question Answering System

Over the last twenty years, significant progress has been made in designing and implementing Question Answering (QA) systems. However, addressing complex questions, the answers to which are spread across multiple documents, remains a challenging problem. Recent QA systems that are designed to handle complex questions work either on the basis of knowledge graphs, or utilise contem- porary neural models that are expensive to train, in terms of both computational resources and the volume of training data required. In this paper, we present LiCQA, an unsupervised question answer- ing model that works primarily on the basis of corpus evidence. We empirically compare the effectiveness and efficiency of LiCQA with two recently presented QA systems, which are based on different underlying principles. The results of our experiments show that LiCQA significantly outperforms these two state-of-the-art systems on benchmark data with noteworthy reduction in latency.

cs.CL↗

arXiv2026

Agile Reinforcement Learning through Separable Neural Architecture and Applications

Deep reinforcement learning (RL) is increasingly deployed in resource-constrained environments, yet go-to function approximators - multilayer perceptrons (MLPs) - are often parameter-inefficient due to an imperfect inductive bias for the smooth structure of many value functions. This mismatch can also hinder sample efficiency and slow policy learning in this capacity-limited regime. Although model compression techniques exist, they operate post-hoc and do not improve learning efficiency. Spline-based architectures such as Kolmogorov-Arnold Networks (KANs) have been shown to offer parameter efficiency but are widely reported to exhibit significant computational overhead, especially at scale. In seeking to address these limitations, this work introduces SPAN (SPline-based Adaptive Networks) for RL. SPAN adapts the KHRONOS framework with a learnable preprocessing layer. SPAN is evaluated across discrete (PPO) and high-dimensional continuous (SAC) control tasks, offline settings (Minari/D4RL) and a real-world datacenter HVAC control application. SPAN achieves a 30-50% improvement in sample efficiency and 1.3-9 times higher success rates across benchmarks compared to MLP baselines. Despite incurring a per-step evaluation overhead of 1.2-1.8x, SPAN's superior convergence reliability yields an expected total training cost 1.3-6.3x lower than MLP baselines when accounting for convergence failures. In the HVAC application, SPAN reduces energy consumption in 9 of 12 months relative to MLP while simultaneously achieving a 1.1-3.4x reduction in thermal comfort violations across the evaluation year, demonstrating generalization to real-world engineering control. Furthermore, SPAN demonstrates superior anytime performance and robustness to hyperparameter variations, suggesting it as a viable, high-performance alternative for learning efficient policies in resource-limited settings.

cs.LG↗

arXiv2026

Leviathan: Decoupling Input and Output Representations in Language Models

Modern language models use a single matrix for input embedding and output projection. This couples two distinct objectives: token representation and discrimination over a vocabulary. This work introduces Leviathan, a Transformer architecture that replaces the input embedding matrix with learned embedding vectorization (LEV), a compact continuous mapping from token indices to embeddings. Leviathan's output head remains untied for a parameter increase of as low as 0.2%. Under controlled comparisons with identical Transformer backbones, Leviathan consistently improves language modeling performance over standard tied-embedding baselines across a 200M-1.2B parameter regime on The Pile with gains that grow during training. At 1.2B scale, Leviathan reduces validation perplexity by 9%, requires $2.1\times$ fewer training tokens to reach the tied baseline's final loss, and improves on all six downstream benchmarks evaluated, including a 30% reduction in LAMBADA perplexity. Frequency-stratified analysis reveals gains to be concentrated in rare tokens, where continuous parameterization reduces perplexity by 81%, falling to near zero for the most frequent.

cs.CL↗

arXiv2026

One Word is Enough: Minimal Adversarial Perturbations for Neural Text Ranking

Neural ranking models (NRMs) achieve strong retrieval effectiveness, yet prior work has shown they are vulnerable to adversarial perturbations. We revisit this robustness question with a minimal, query-aware attack that promotes a target document by inserting or substituting a single, semantically aligned word - the query center. We study heuristic and gradient-guided variants, including a white-box method that identifies influential insertion points. On TREC-DL 2019/2020 with BERT and monoT5 re-rankers, our single-word attacks achieve up to 91% success while modifying fewer than two tokens per document on average, achieving competitive rank and score boosts with far fewer edits under a comparable white-box setup to ensure fair evaluation against PRADA. We also introduce new diagnostic metrics to analyze attack sensitivity beyond aggregate success rates. Our analysis reveals a Goldilocks zone in which mid-ranked documents are most vulnerable. These findings demonstrate practical risks and motivate future defenses for robust neural ranking.

cs.IR↗

arXiv2026

LLMs as Assessors: Right for the Right Reason?

A good deal of recent research has focused on how Large Language Models (LLMs) may be used as judges in place of humans to evaluate the quality of the output produced by various text / image processing systems. Within this broader context, a number of studies have investigated the specific question of how effectively LLMs can be used as relevance assessors for the standard ad hoc task in Information Retrieval (IR). We extend these studies by looking at additional questions. Most importantly, we use a Wikipedia based test collection created by the INEX initiative, and prompt LLMs to not only judge whether documents are relevant / non-relevant, but to highlight relevant passages in documents that it regards as useful. The human relevance assessors involved in creating this collection were given analogous instructions, i.e., they were asked to highlight all passages within a document that respond to the information need expressed in a query. This enables us to evaluate the quality of LLMs as judges not only at the document level, but to also quantify how often these judges are right for the right reasons. Our observations lead us to reiterate the cautionary note sounded in some earlier studies when it comes to using LLMs as assessors for creating IR datasets: while LLMs are unquestionably promising, and may be used judiciously to subtantially reduce the amount of human involvement required to generate high-quality benchmark datasets, they cannot replace humans as assessors.

cs.IR↗

arXiv2025

A Kernel-based Resource-efficient Neural Surrogate for Multi-fidelity Prediction of Aerodynamic Field

Surrogate models provide fast alternatives to costly aerodynamic simulations and are extremely useful in design and optimization applications. This study proposes the use of a recent kernel-based neural surrogate, KHRONOS. In this work, we blend sparse high-fidelity (HF) data with low-fidelity (LF) information to predict aerodynamic fields under varying constraints in computational resources. Unlike traditional approaches, KHRONOS is built upon variational principles, interpolation theory, and tensor decomposition. These elements provide a mathematical basis for heavy pruning compared to dense neural networks. Using the AirfRANS dataset as a high-fidelity benchmark and NeuralFoil to generate low-fidelity counterparts, this work compares the performance of KHRONOS with three contemporary model architectures: a multilayer perceptron (MLP), a graph neural network (GNN), and a physics-informed neural network (PINN). We consider varying levels of high-fidelity data availability (0%, 10%, and 30%) and increasingly complex geometry parameterizations. These are used to predict the surface pressure coefficient distribution over the airfoil. Results indicate that, whilst all models eventually achieve comparable predictive accuracy, KHRONOS excels in resource-constrained conditions. In this domain, KHRONOS consistently requires orders of magnitude fewer trainable parameters and delivers much faster training and inference than contemporary dense neural networks at comparable accuracy. These findings highlight the potential of KHRONOS and similar architectures to balance accuracy and efficiency in multi-fidelity aerodynamic field prediction.

cs.LG↗

arXiv2025

A Unified Generative-Predictive Framework for Deterministic Inverse Design

Inverse design of heterogeneous material microstructures is a fundamentally ill-posed and famously computationally expensive problem. This is exacerbated by the high-dimensional design spaces associated with finely resolved images, multimodal input property streams, and a highly nonlinear forward physics. Whilst modern generative models excel at accurately modeling such complex forward behavior, most of them are not intrinsically structured to support fast, stable \emph{deterministic} inversion with a physics-informed bias. This work introduces Janus, a unified generative-predictive framework to address this problem. Janus couples a deep encoder-decoder architecture with a predictive KHRONOS head, a separable neural architecture. Topologically speaking, Janus learns a latent manifold simultaneously isometric for generative inversion and pruned for physical prediction; the joint objective inducing \emph{disentanglement} of the latent space. Janus is first validated on the MNIST dataset, demonstrating high-fidelity reconstruction, accurate classification and diverse generative inversion of all ten target classes. It is then applied to the inverse design of heterogeneous microstructures labeled with thermal conductivity. It achieves a forward prediction accuracy $R^2=0.98$ (2\% relative error) and sub-5\% pixelwise reconstruction error. Inverse solutions satisfy target properties to within $1\%$ relative error. Inverting a sweep through properties reveal smooth traversal of the latent manifold, and UMAP visualization confirms the emergence of a low-dimensional, disentangled manifold. By unifying prediction and generation within a single latent space, Janus enables real-time, physics-informed inverse microstructure generation at a lower computational cost typically associated with classical optimization-based approaches.

cs.LG↗

arXiv2025

Process Microstructure Coupling in Reduced Gravity Laser Welding via Open-Source Multiphysics Simulation Framework

Supplying spare parts from Earth for in space repair is economically prohibitive and logistically slow, posing a major barrier to sustainable space operations. As lunar and Martian missions accelerate in the coming decades, the feasibility of in-situ repair methods, particularly laser based welding, must be rigorously evaluated. The micro scale physics governing weld quality are fundamentally altered by variations in gravity and ambient pressure, yet their coupled influence across different welding regimes remains poorly understood. This work introduces a fully open-source thermo-fluid-microstructure modeling framework with computational fluid dynamics (CFD) and cellular automata (CA) to quantify how gravitational conditions reshape weld pool behavior across multiple welding regimes and spatial scales. The framework further enables prediction of the resulting microstructure from imposed process parameters, providing an accessible pathway for future research in in space manufacturing (ISM). As a demonstration, the article analyzes laser welding of Al6061 across three process regimes (conduction, transition, and keyhole) and four gravitational environments (Earth, Moon, Mars, and International Space Station (ISS)). Analysis reveals that reduced gravity suppresses buoyancy-driven convection, thereby altering melt pool geometries. Vacuum conditions increase laser energy deposition, while microgravity promotes equiaxed grain formation via reduced melt convection. The framework captures keyhole dynamics, thermal histories, and grain morphologies with high fidelity, validated against experimental data. These findings establish process microstructure relationships critical for reliable metallic fabrication in space and provide a reproducible platform for future in-space manufacturing research.

physics.flu-dyn↗

arXiv2025

Retriv at BLP-2025 Task 1: A Transformer Ensemble and Multi-Task Learning Approach for Bangla Hate Speech Identification

This paper addresses the problem of Bangla hate speech identification, a socially impactful yet linguistically challenging task. As part of the "Bangla Multi-task Hate Speech Identification" shared task at the BLP Workshop, IJCNLP-AACL 2025, our team "Retriv" participated in all three subtasks: (1A) hate type classification, (1B) target group identification, and (1C) joint detection of type, severity, and target. For subtasks 1A and 1B, we employed a soft-voting ensemble of transformer models (BanglaBERT, MuRIL, IndicBERTv2). For subtask 1C, we trained three multitask variants and aggregated their predictions through a weighted voting ensemble. Our systems achieved micro-f1 scores of 72.75% (1A) and 72.69% (1B), and a weighted micro-f1 score of 72.62% (1C). On the shared task leaderboard, these corresponded to 9th, 10th, and 7th positions, respectively. These results highlight the promise of transformer ensembles and weighted multitask frameworks for advancing Bangla hate speech detection in low-resource contexts. We made experimental scripts publicly available for the community.

cs.CL↗

arXiv2025

Retriv at BLP-2025 Task 2: Test-Driven Feedback-Guided Framework for Bangla-to-Python Code Generation

Large Language Models (LLMs) have advanced the automated generation of code from natural language prompts. However, low-resource languages (LRLs) like Bangla remain underrepresented due to the limited availability of instruction-to-code datasets and evaluation benchmarks. To address this, the BLP Workshop at IJCNLP-AACL 2025 introduced a shared task on "Code Generation in Bangla". In this work, we propose a method that combines instruction prompting with a test-driven, feedback-guided iterative refinement process using a fine-tuned Qwen2.5-14B model. The model generates code from Bangla instructions, tests it against unit tests, and iteratively refines any failing outputs through three evaluation passes, using test feedback to guide each step. This approach helped our team "Retriv" to secure 2nd place in the shared task with a Pass@1 score of 0.934. The analysis highlights challenges in Bangla instruction understanding and Python code generation, emphasizing the need for targeted methods in LRLs. We made experimental scripts publicly available for the community.

cs.CL↗

arXiv2025

A Liquid-Fueled Reactor Network Model for Enhanced NOx Prediction in Gas Turbine Combustors

This study introduces a liquid-fueled reactor network (LFRN) framework for reduced-order modeling of gas turbine combustors. The proposed LFRN extends conventional gaseous-fueled reactor network methods by incorporating specialized reactors that account for spray breakup, droplet heating, and evaporation, thereby enabling the treatment of multiphase effects essential to liquid-fueled systems. Validation is performed against detailed computational fluid dynamics (CFD) simulations of a liquid-fueled can combustor, with parametric studies conducted across variations in inlet air temperature and fuel flow rate. Results show that the LFRN substantially reduces NOx prediction errors relative to gaseous reactor networks while maintaining accurate outlet temperature predictions. A sensitivity analysis on the number of clusters demonstrates progressive convergence toward the CFD predictions with increasing network complexity. In terms of computational efficiency, the LFRN achieves runtimes of $O(10s)$ on a single CPU core, representing speed-ups generally exceeding 1,000$\times$ compared to CFD. Overall, the findings demonstrate the potential of the LFRN as a computationally efficient reduced-order modeling tool that complements CFD to enable rapid emissions assessment and design-space exploration for liquid-fueled gas turbine combustors.

physics.flu-dyn↗

arXiv2025

The Method of Infinite Descent

Training - the optimisation of complex models - is traditionally performed through small, local, iterative updates [D. E. Rumelhart, G. E. Hinton, R. J. Williams, Nature 323, 533-536 (1986)]. Approximating solutions through truncated gradients is a paradigm dating back to Cauchy [A.-L. Cauchy, Comptes Rendus Math\'ematique 25, 536-538 (1847)] and Newton [I. Newton, The Method of Fluxions and Infinite Series (Henry Woodfall, London, 1736)]. This work introduces the Method of Infinite Descent, a semi-analytic optimisation paradigm that reformulates training as the direct solution to the first-order optimality condition. By analytical resummation of its Taylor expansion, this method yields an exact, algebraic equation for the update step. Realisation of the infinite Taylor tower's cascading resummation is formally derived, and an exploitative algorithm for the direct solve step is proposed. This principle is demonstrated with the herein-introduced AION (Analytic, Infinitely-Optimisable Network) architecture. AION is a model designed expressly to satisfy the algebraic closure required by Infinite Descent. In a simple test problem, AION reaches the optimum in a single descent step. Together, this optimiser-model pair exemplify how analytic structure enables exact, non-iterative convergence. Infinite Descent extends beyond this example, applying to any appropriately closed architecture. This suggests a new class of semi-analytically optimisable models: the \emph{Infinity Class}; sufficient conditions for class membership are discussed. This offers a pathway toward non-iterative learning.

cs.LG↗

arXiv2025

Gravity and Composition Modulated Solidification and Mechanical Properties of Al-Cu Nanostructures

The future of space exploration and human settlement beyond Earth hinges on a deeper understanding of in space manufacturing processes. The unique physical conditions and scarcity of experimental data demand robust computational models to investigate the atomic scale physics of solidification. This work presents a molecular dynamics (MD) model to examine the solidification behavior of the Al Cu binary alloy, focusing on the influence of varying compositions and gravity levels (Earth, Lunar, Martian, and microgravity) on atomistic solidification mechanisms and the resulting mechanical properties specifically, hardness of as solidified nanostructures. Hardness is evaluated via nanoindentation simulations. The study confirms that gravitational forces significantly affect the solidification pathways of Al Cu alloys. Notably, by tuning alloy composition, the influence of gravity can be modulated and in some cases, even reversed. Moreover, hardness exhibits a coupled dependence on both composition and gravity, offering a promising avenue for bottom-up design of components tailored for extraterrestrial environments. The article delves into the nanoscale physical mechanisms underlying these phenomena and outlines future directions for extending this modeling framework to broader applications.

cond-mat.mtrl-sci↗

arXiv2025

Explainable Hierarchical Deep Learning Neural Networks (Ex-HiDeNN)

Data-driven science and computation have advanced immensely to construct complex functional relationships using trainable parameters. However, efficiently discovering interpretable and accurate closed-form expressions from complex dataset remains a challenge. The article presents a novel approach called Explainable Hierarchical Deep Learning Neural Networks or Ex-HiDeNN that uses an accurate, frugal, fast, separable, and scalable neural architecture with symbolic regression to discover closed-form expressions from limited observation. The article presents the two-step Ex-HiDeNN algorithm with a separability checker embedded in it. The accuracy and efficiency of Ex-HiDeNN are tested on several benchmark problems, including discerning a dynamical system from data, and the outcomes are reported. Ex-HiDeNN generally shows outstanding approximation capability in these benchmarks, producing orders of magnitude smaller errors compared to reference data and traditional symbolic regression. Later, Ex-HiDeNN is applied to three engineering applications: a) discovering a closed-form fatigue equation, b) identification of hardness from micro-indentation test data, and c) discovering the expression for the yield surface with data. In every case, Ex-HiDeNN outperformed the reference methods used in the literature. The proposed method is built upon the foundation and published works of the authors on Hierarchical Deep Learning Neural Network (HiDeNN) and Convolutional HiDeNN. The article also provides a clear idea about the current limitations and future extensions of Ex-HiDeNN.

cs.LG↗

arXiv2025

INN-FF: A Scalable and Efficient Machine Learning Potential for Molecular Dynamics

The ability to accurately model interatomic interactions in large-scale systems is fundamental to understanding a wide range of physical and chemical phenomena, from drug-protein binding to the behavior of next-generation materials. While machine learning interatomic potentials (MLIPs) have made it possible to achieve ab initio-level accuracy at significantly reduced computational cost, they still require very large training datasets and incur substantial training time and expense. In this work, we propose the Interpolating Neural Network Force Field (INN-FF), a novel framework that merges interpolation theory and tensor decomposition with neural network architectures to efficiently construct molecular dynamics potentials from limited quantum mechanical data. Interpolating Neural Networks (INNs) achieve comparable or better accuracy than traditional multilayer perceptrons (MLPs) while requiring orders of magnitude fewer trainable parameters. On benchmark datasets such as liquid water and rMD17, INN-FF not only matches but often surpasses state-of-the-art accuracy by an order of magnitude, while achieving significantly lower error when trained on smaller datasets. These results suggest that INN-FF offers a promising path toward building efficient and scalable machine-learned force fields.

cond-mat.mtrl-sci↗