Partha Pratim Chakrabarti

Tanmoy Chakraborty, Ayan Sengupta, Suparna Bhattacharya et al.

Large language models (LLMs) frequently achieve impressive scores on standardized benchmarks, yet accuracy alone offers a limited view of their capabilities. Evaluating open-source LLMs through leaderboards faces persistent issues like data contamination, narrow task scope, and weak alignment with real-world reliability. Benchmark-based evaluations such as MMLU PRO, BBH, or IFEval primarily capture \textit{what} a model outputs on fixed test sets, not \textit{how} it processes information, calibrates uncertainty, or structures internal knowledge. In this article, we advocate for a shift from benchmark-centric evaluation toward a complementary, \textit{state-centered intrinsic assessment} of LLMs. To this end, we introduce \textbf{Latent Performance Profiling (LPP)} -- a framework that derives task-agnostic diagnostics from hidden activations and output distributions. LPP defines a set of scalar metrics on a model's latent representations and dynamics, revealing scale-independent traits that enable interpretable comparisons and uncover hidden vulnerabilities. Unlike static accuracy scores, LPP provides stable, architecture-sensitive signatures across models of similar size. With extensive empirical analyses across eight LLMs, spanning a size range of 0.5B-14B, we demonstrate that models with similar benchmark scores can exhibit contrasting latent profiles, such as differences in entropy or adaptability. Guided by these insights, we design synthetic probes for uncertainty and symbolic reasoning that align with intrinsic metrics while decoupling from leaderboard bias. We recommend that reporting LPP alongside benchmarks provides a deeper, interpretable understanding of model behavior, enabling more reliable model selection, safety assessment, and evaluation beyond surface-level accuracy.

Manaar Alam, Shubhajit Datta, Debdeep Mukhopadhyay et al.

The security of deep learning (DL) systems is an extremely important field of study as they are being deployed in several applications due to their ever-improving performance to solve challenging tasks. Despite overwhelming promises, the deep learning systems are vulnerable to crafted adversarial examples, which may be imperceptible to the human eye, but can lead the model to misclassify. Protections against adversarial perturbations on ensemble-based techniques have either been shown to be vulnerable to stronger adversaries or shown to lack an end-to-end evaluation. In this paper, we attempt to develop a new ensemble-based solution that constructs defender models with diverse decision boundaries with respect to the original model. The ensemble of classifiers constructed by (1) transformation of the input by a method called Split-and-Shuffle, and (2) restricting the significant features by a method called Contrast-Significant-Features are shown to result in diverse gradients with respect to adversarial attacks, which reduces the chance of transferring adversarial examples from the original to the defender model targeting the same class. We present extensive experimentations using standard image classification datasets, namely MNIST, CIFAR-10 and CIFAR-100 against state-of-the-art adversarial attacks to demonstrate the robustness of the proposed ensemble-based defense. We also evaluate the robustness in the presence of a stronger adversary targeting all the models within the ensemble simultaneously. Results for the overall false positives and false negatives have been furnished to estimate the overall performance of the proposed methodology.

Aritra Mitra, Srijoni Majumdar, Anamitra Mukhopadhyay et al.

Comments are very useful to the flow of code development. With the increasing commonality of code, novice coders have been creating a significant amount of codebases. Due to lack of commenting standards, their comments are often useless, and increase the time taken to further maintain codes. This study intends to find the usefulness of large language models (LLMs) in these cases to generate potentially better comments. This study focuses on the feasibility of design documents as a context for the LLMs to generate more useful comments, as design documents are often used by maintainers to understand code when comments do not suffice.

Dvij Kalaria, Aritra Hazra, Partha Pratim Chakrabarti

With the rapid advancement and increased use of deep learning models in image identification, security becomes a major concern to their deployment in safety-critical systems. Since the accuracy and robustness of deep learning models are primarily attributed from the purity of the training samples, therefore the deep learning architectures are often susceptible to adversarial attacks. Adversarial attacks are often obtained by making subtle perturbations to normal images, which are mostly imperceptible to humans, but can seriously confuse the state-of-the-art machine learning models. We propose a framework, named APuDAE, leveraging Denoising AutoEncoders (DAEs) to purify these samples by using them in an adaptive way and thus improve the classification accuracy of the target classifier networks that have been attacked. We also show how using DAEs adaptively instead of using them directly, improves classification accuracy further and is more robust to the possibility of designing adaptive attacks to fool them. We demonstrate our results over MNIST, CIFAR-10, ImageNet dataset and show how our framework (APuDAE) provides comparable and in most cases better performance to the baseline methods in purifying adversaries. We also design adaptive attack specifically designed to attack our purifying model and demonstrate how our defense is robust to that.

Varun Madhavan, Adway Mitra, Partha Pratim Chakrabarti

Agent-Based Models are very useful for simulation of physical or social processes, such as the spreading of a pandemic in a city. Such models proceed by specifying the behavior of individuals (agents) and their interactions, and parameterizing the process of infection based on such interactions based on the geography and demography of the city. However, such models are computationally very expensive, and the complexity is often linear in the total number of agents. This seriously limits the usage of such models for simulations, which often have to be run hundreds of times for policy planning and even model parameter estimation. An alternative is to develop an emulator, a surrogate model that can predict the Agent-Based Simulator's output based on its initial conditions and parameters. In this paper, we discuss a Deep Learning model based on Dilated Convolutional Neural Network that can emulate such an agent based model with high accuracy. We show that use of this model instead of the original Agent-Based Model provides us major gains in the speed of simulations, allowing much quicker calibration to observations, and more extensive scenario analysis. The models we consider are spatially explicit, as the locations of the infected individuals are simulated instead of the gross counts. Another aspect of our emulation framework is its divide-and-conquer approach that divides the city into several small overlapping blocks and carries out the emulation in them parallelly, after which these results are merged together. This ensures that the same emulator can work for a city of any size, and also provides significant improvement of time complexity of the emulator, compared to the original simulator.

Ganesh Sahadeo Meshram, Suman Chakraborty, Nishant Sinha et al.

In the complex domain of microfluidics systems, analysing fluid flow patterns through random-shaped circular microchannels is significantly challenging task. Conventional approach of solving such problems using computational fluid dynamics often incapable due to their intensive computational requirements and high simulation times. In this study, addressing these limitations, we introduce $μ$-FlowNet, a deep learning framework based on the adaptable U-Net autoencoders. This model provides a data-driven approach that enhances the prediction and mapping of random-shaped circular microchannels and their corresponding fluid flow patterns. The datasets required for the training of the model is generated by performing extensive simulations using conventional approach of computational fluid dynamics methods. The datasets are then pre-processed and accessed the required spatial and temporal features that are essential for the training. We have trained three different models based on U-Net framework namely, standard U-Net, T-Net, and U-Net with attention mechanism to compare the prediction accuracy and loss. The accuracy of the $μ$-FlowNet is compared using metrics of dice score and intersection over union and it shows that U-Net with attention mechanism shows the highest dice score and IoU of 0.9317 and 0.8731, respectively and shows the highest structural similarity as compared to standard U-Net and T-Net. This show that U-Net with attention mechanism serves best model to map the fluid flow pattern with random datasets on testing.

Ganesh Sahadeo Meshram, Partha Pratim Chakrabarti, Suman Chakraborty

Spreading of liquid droplets on solid substrates constitutes a classic multiphysics problem with widespread applications ranging from inkjet printing, spray cooling, to biomedical microfluidic systems. Yet, accurate computational fluid dynamic (CFD) simulations are prohibitively expensive, taking more than 18 to 24 hours for each transient computation. In this paper, Physics-Informed Laplace Operator Neural Network (PI-LNO) is introduced, representing a novel architecture where the Laplace integral transform function serves as a learned physics-informed functional basis. Extensive comparative benchmark studies were performed against five other state-of-the-art approaches: UNet, UNet with attention modules (UNet-AM), DeepONet, Physics-Informed UNet (PI-UNet), and Laplace Neural Operator (LNO). Through complex Laplace transforms, PI-LNO natively models the exponential transient dynamics of the spreading process. A TensorFlow-based PI-LNO is trained on multi-surface CFD data spanning contact angles $θ_s ε[20,160]$, employing a physics-regularized composite loss combining data fidelity (MSE, MAE, RMSE) with Navier-Stokes, Cahn-Hilliard, and causality constraints.

Ganesh Sahadeo Meshram, Partha Pratim Chakrabarti, Suman Chakraborty

We introduce a Lattice-Boltzmann-driven kinetic physics-informed neural network (K-PINN) for predictive modeling of droplet dynamics on structured surfaces, in which the discrete Boltzmann-BGK equation is incorporated into the learning framework. Different from traditional PINNs that are restricted by macroscopic continuum equations, the K-PINN framework is built on the mesoscopic kinetic level, in which the essential Lattice-Boltzmann physics is preserved in the data-efficient neural network. The K-PINN has been successfully employed for modeling non-trivial droplet phenomena such as contact pinning, anisotropic spreading, and capillary hysteresis on substrates of different morphologies, ranging from random roughness to periodic pillar structures. Moreover, strict physical consistency, such as mass conservation within 1.5%, is ensured in the K-PINN framework. Furthermore, the U-Net-based encoder-decoder structure of the K-PINN results in a 50-75% reduction in error compared to traditional neural networks, achieving almost perfect agreement with high-resolution Lattice-Boltzmann simulations $L_2$ ~ 0.021-0.026, $R^2$ ~ 0.999. Robust convergence of the K-PINN to diverse surface morphologies is ensured through curriculum learning and adaptive two-phase optimization. Upon convergence, the K-PINN can perform real-time prediction with over 104 evaluations per second. Through the combination of kinetic theory and physics-informed learning, this work establishes a new paradigm for fast, physically consistent modeling of multiphase flows on complex surfaces.

Ganesh Sahadeo Meshram, Partha Pratim Chakrabarti, Suman Chakraborty

In this study, we have explored an application of deep learning architecture of the U-Net model, originally designed for biomedical image segmentation, in a regression analysis aimed at predicting fluid flows through textured microchannels. The data for this analysis is generated using the lattice Boltzmann method through extensive simulations, capturing the intricate behaviors of fluid dynamics in a microscale environment. The raw simulation data was meticulously preprocessed to prepare it for training the U-Net model, ensuring that the input features and labels were appropriately formatted and normalized to optimize the learning process of the model. The U-Net model, with its inherent capability of capturing spatial hierarchies and producing better predictions, proved effective in this novel application. We have evaluated the performance of the model using metrics including MSE, RMSE, MAE, and $R^2$ scores. These metrics were crucial in assessing the accuracy and reliability of the model predictions. The results demonstrate that the U-Net model can predict fluid flows with high accuracy and less error, indicating its potential for broader applications in fluid dynamics and other fields requiring precise regression modeling. A parametric analysis of the U-Net with attention mechanism showed that the velocity field prediction is contingent upon the solid-fluid interaction parameter and surface wettability. The U-Net equipped with an attention mechanism predicts the velocity magnitude and components for textured microchannels with an average error of 5.18%, which upon optimization may subsequently lower to 2.1%. The U-Net model including an attention mechanism (U-Net AM) regularly surpasses the conventional U-Net model in all measures, evidencing enhanced accuracy and generalization.

Ganesh Sahadeo Meshram, Partha Pratim Chakrabarti, Suman Chakraborty

One of the biggest challenges in the optimization of micro-scale fluid transport phenomena is the prediction of unsteady fluid flow in the presence of rough channel walls. Even though the accuracy of available computational fluid dynamics (CFD) solvers such as the lattice Boltzmann method (LBM) is satisfactory, the computational cost of design exploration is very high due to the diverse range of geometries and flow regimes involved in microchannel flows. The present paper introduces a revolutionary concept of a ground-breaking physics-informed neural network (PINN) that utilizes sparse lattice Boltzmann data in combination with the Navier-Stokes equations for the prediction of unsteady fluid flow in fractal-rough microchannels. The roughness of the channel walls is represented by the Weierstrass-Mandelbrot function, considering the characteristics of the surface roughness in real-life problems. The constraints of the Navier-Stokes equations are incorporated in the loss function of the PINN concept for achieving accuracy at much lower computational costs of 150-200 times fewer data points. The validation of the accuracy of the reconstruction of the flow fields is carried out for different Reynolds numbers ranging from Re = 1 to 45 and different amplitude values of the rough channel walls ranging from 5 to 20 lattice units.

Alapan Kuila, Somnath Jena, Sudeshna Sarkar et al.

In today's media landscape, where news outlets play a pivotal role in shaping public opinion, it is imperative to address the issue of sentiment manipulation within news text. News writers often inject their own biases and emotional language, which can distort the objectivity of reporting. This paper introduces a novel approach to tackle this problem by reducing the polarity of latent sentiments in news content. Drawing inspiration from adversarial attack-based sentence perturbation techniques and a prompt based method using ChatGPT, we employ transformation constraints to modify sentences while preserving their core semantics. Using three perturbation methods: replacement, insertion, and deletion coupled with a context-aware masked language model, we aim to maximize the desired sentiment score for targeted news aspects through a beam search algorithm. Our experiments and human evaluations demonstrate the effectiveness of these two models in achieving reduced sentiment polarity with minimal modifications while maintaining textual similarity, fluency, and grammatical correctness. Comparative analysis confirms the competitive performance of the adversarial attack based perturbation methods and prompt-based methods, offering a promising solution to foster more objective news reporting and combat emotional language bias in the media.

Bhavesh Neekhra, Debayan Gupta, Partha Pratim Chakrabarti

Feature selection (FS) is assumed to improve predictive performance and identify meaningful features in high-dimensional datasets. Surprisingly, small random subsets of features (0.02-1%) match or outperform the predictive performance of both full feature sets and FS across 28 out of 30 diverse datasets (microarray, bulk and single-cell RNA-Seq, mass spectrometry, imaging, etc.). In short, any arbitrary set of features is as good as any other (with surprisingly low variance in results) - so how can a particular set of selected features be "important" if they perform no better than an arbitrary set? These results challenge the assumption that computationally selected features reliably capture meaningful signals, emphasizing the importance of rigorous validation before interpreting selected features as actionable, particularly in computational genomics.

Debjyoti Das Adhikary, Aritra Hazra, Partha Pratim Chakrabarti

Image explanation has been one of the key research interests in the Deep Learning field. Throughout the years, several approaches have been adopted to explain an input image fed by the user. From detecting an object in a given image to explaining it in human understandable sentence, to having a conversation describing the image, this problem has seen an immense change throughout the years, However, the existing works have been often found to (a) hallucinate objects that do not exist in the image and/or (b) lack identifying the complete set of objects present in the image. In this paper, we propose a novel approach to mitigate these drawbacks of inconsistency and incompleteness of the objects recognized during the image explanation. To enable this, we propose an interpretable framework that can be plugged atop diverse image explaining frameworks including Image Captioning, Visual Question Answering (VQA) and Prompt-based AI using LLMs, thereby enhancing their explanation capabilities by rectifying the incorrect or missing objects. We further measure the efficacy of the rectified explanations generated through our proposed approaches leveraging object based precision metrics, and showcase the improvements in the inconsistency and completeness of image explanations. Quantitatively, the proposed framework is able to improve the explanations over the baseline architectures of Image Captioning (improving the completeness by 81.81% and inconsistency by 37.10%), Visual Question Answering(average of 9.6% and 37.10% in completeness and inconsistency respectively) and Prompt-based AI model (0.01% and 5.2% for completeness and inconsistency respectively) surpassing the current state-of-the-art by a substantial margin.

Kushal Kedia, Rajat Kumar Jenamani, Aritra Hazra et al.

Multi-Agent Path Finding (MAPF) is the problem of finding collision-free paths for multiple agents from their start locations to end locations. We consider an extension to this problem, Precedence Constrained Multi-Agent Path Finding (PC-MAPF), wherein agents are assigned a sequence of planning tasks that contain precedence constraints between them. PC-MAPF has various applications, for example in multi-agent pickup and delivery problems where some objects might require multiple agents to collaboratively pickup and move them in unison. Precedence constraints also arise in warehouse assembly problems where before a manufacturing task can begin, its input resources must be manufactured and delivered. We propose a novel algorithm, Precedence Constrained Conflict Based Search (PC-CBS), which finds makespan-optimal solutions for this class of problems. PC-CBS utilizes a Precedence-Constrained Task-Graph to define valid intervals for each planning task and updates them when precedence conflicts are encountered. We benchmark the performance of this algorithm over various warehouse assembly, and multi-agent pickup and delivery tasks, and use it to evaluate the sub-optimality of a recently proposed efficient baseline.

Shubhi Shukla, Subhadeep Dalui, Manaar Alam et al.

Adversarial attacks rely on transferability, where an adversarial example (AE) crafted on a surrogate classifier tends to mislead a target classifier. Recent ensemble methods demonstrate that AEs are less likely to mislead multiple classifiers in an ensemble. This paper proposes a new ensemble training using a Pairwise Adversarially Robust Loss (PARL) that by construction produces an ensemble of classifiers with diverse decision boundaries. PARL utilizes outputs and gradients of each layer with respect to network parameters in every classifier within the ensemble simultaneously. PARL is demonstrated to achieve higher robustness against black-box transfer attacks than previous ensemble methods as well as adversarial training without adversely affecting clean example accuracy. Extensive experiments using standard Resnet20, WideResnet28-10 classifiers demonstrate the robustness of PARL against state-of-the-art adversarial attacks. While maintaining similar clean accuracy and lesser training time, the proposed architecture has a 24.8% increase in robust accuracy ($ε$ = 0.07) from the state-of-the art method.

Dvij Kalaria, Aritra Hazra, Partha Pratim Chakrabarti

With the rapid advancement and increased use of deep learning models in image identification, security becomes a major concern to their deployment in safety-critical systems. Since the accuracy and robustness of deep learning models are primarily attributed from the purity of the training samples, therefore the deep learning architectures are often susceptible to adversarial attacks. Adversarial attacks are often obtained by making subtle perturbations to normal images, which are mostly imperceptible to humans, but can seriously confuse the state-of-the-art machine learning models. What is so special in the slightest intelligent perturbations or noise additions over normal images that it leads to catastrophic classifications by the deep neural networks? Using statistical hypothesis testing, we find that Conditional Variational AutoEncoders (CVAE) are surprisingly good at detecting imperceptible image perturbations. In this paper, we show how CVAEs can be effectively used to detect adversarial attacks on image classification networks. We demonstrate our results over MNIST, CIFAR-10 dataset and show how our method gives comparable performance to the state-of-the-art methods in detecting adversaries while not getting confused with noisy images, where most of the existing methods falter.

Ginju V. George, Aritra Hazra, Pallab Dasgupta et al.

Reliability and fault tolerance are critical attributes of embedded cyber-physical systems that require a high safety-integrity level. For such systems, the use of formal functional safety specifications has been strongly advocated in most industrial safety standards, but reliability and fault tolerance have traditionally been treated as platform issues. We believe that addressing reliability and fault tolerance at the functional safety level widens the scope for resource optimization, targeting those functionalities that are safety-critical, rather than the entire platform. Moreover, for software based control functionalities, temporal redundancies have become just as important as replication of physical resources, and such redundancies can be modeled at the functional specification level. The ability to formally model functional reliability at a specification level enables early estimation of physical resources and computation bandwidth requirements. In this paper we propose, for the first time, a resource estimation methodology from a formal functional safety specification augmented by reliability annotations. The proposed reliability specification is overlaid on the safety-critical functional specification and our methodology extracts a constraint satisfaction problem for determining the optimal set of resources for meeting the reliability target for the safety-critical behaviors. We use SMT (Satisfiability Modulo Theories) / ILP (Integer Linear Programming) solvers at the back end to solve the optimization problem, and demonstrate the feasibility of our methodology on a Satellite Launch Vehicle Navigation, Guidance and Control (NGC) System.

Rajorshee Raha, Soumyajit Dey, Partha Pratim Chakrabarti et al.

Recent studies have shown that adaptively regulating the sampling rate results in significant reduction in computational resources in embedded software based control. Selecting a uniform sampling rate for a control loop is robust, but overtly pessimistic for sharing processors among multiple control loops. Fine grained regulation of periodicity achieves better resource utilization, but is hard to implement online in a robust way. In this paper we propose multi-mode sampling period selection, derived from an offline control theoretic analysis of the system. We report significant gains in computational efficiency without trading off control performance.