arXiv subjects

Rajiv Misra

Publications and source records attributed to Rajiv Misra.

12 recordsLinked to original sources

arXiv2026

Posterior-Calibrated Causal Circuits in Variational Autoencoders: Why Image-Domain Interpretability Fails on Tabular Data

Although mechanism-based interpretability has generated an abundance of insight for discriminative network analysis, generative models are less understood -- particularly outside of image-related applications. We investigate how much of the causal circuitry found within image-related variational autoencoders (VAEs) will generalize to tabular data, as VAEs are increasingly used for imputation, anomaly detection, and synthetic data generation. In addition to extending a four-level causal intervention framework to four tabular and one image benchmark across five different VAE architectures (with 75 individual training runs per architecture and three random seed values for each run), this paper introduces three new techniques: posterior-calibration of Causal Effect Strength (CES), path-specific activation patching, and Feature-Group Disentanglement (FGD). The results from our experiments demonstrate that: (i) Tabular VAEs have circuits with modularity that is approximately 50% lower than their image counterparts. (ii) $\beta$-VAE experiences nearly complete collapse in CES scores when applied to heterogeneous tabular features (0.043 CES score for tabular data compared to 0.133 CES score for images), which can be directly attributed to reconstruction quality degradation (r = -0.886 correlation coefficient between CES and MSE). (iii) CES successfully captures nine of eleven statistically significant architecture differences using Holm--\v{S}id\'{a}k corrections. (iv) Interventions with high specificity predict the highest downstream AUC values (r = 0.460, p < .001). This study challenges the common assumption that architectural guidance from image-related studies can be transferred to tabular datasets.

cs.LG↗

arXiv2026

Detection Without Correction: A Robust Asymmetry in Activation-Based Hallucination Probing

Activation-based linear probing is widely proposed as a method for both detecting and correcting hallucinations in autoregressive language models. We present an empirical study across seven models spanning 117M to 7B parameters and three architecture families (GPT-2, Pythia, Qwen-2.5) that documents a robust asymmetry: linear probes can detect hallucination signals with above-chance accuracy in larger models, but activation steering along the probe-derived direction fails to correct hallucinations in 7 of 7 models tested. We further find that output-confidence baselines outperform activation probes on raw detection AUC at every model above 410M parameters, with the gap reaching 0.157 AUC for Pythia-6.9B. The probe's distinguishing value is therefore not detection accuracy but temporal positioning: probe signals are accessible at position zero (before any output tokens are produced), enabling pre-generation flagging that output-based methods structurally cannot provide. The temporal signal is statistically significant in two of seven models (Pythia-1.4B, p = 0.012; Qwen2.5-7B, p = 0.038) and absent in models below 400M parameters and in the base-only Pythia-6.9B. We position these findings as a clean negative result for the dominant probing-as-detection-and-control research direction and as initial evidence that probe-based methods occupy a complementary deployment niche, namely pre-generation flagging, rather than competing with output-based detectors on raw accuracy.

cs.CL↗

arXiv2026

Fundamental Limits of Neural Network Sparsification: Evidence from Catastrophic Interpretability Collapse

Extreme neural network sparsification (90% activation reduction) presents a critical challenge for mechanistic interpretability: understanding whether interpretable features survive aggressive compression. This work investigates feature survival under severe capacity constraints in hybrid Variational Autoencoder--Sparse Autoencoder (VAE-SAE) architectures. We introduce an adaptive sparsity scheduling framework that progressively reduces active neurons from 500 to 50 over 50 training epochs, and provide empirical evidence for fundamental limits of the sparsification-interpretability relationship. Testing across two benchmark datasets -- dSprites and Shapes3D -- with both Top-k and L1 sparsification methods, our key finding reveals a pervasive paradox: while global representation quality (measured by Mutual Information Gap) remains stable, local feature interpretability collapses systematically. Under Top-k sparsification, dead neuron rates reach $34.4\pm0.9\%$ on dSprites and $62.7\pm1.3\%$ on Shapes3D at k=50. L1 regularization -- a fundamentally different "soft constraint" paradigm -- produces equal or worse collapse: $41.7\pm4.4\%$ on dSprites and $90.6\pm0.5\%$ on Shapes3D. Extended training for 100 additional epochs fails to recover dead neurons, and the collapse pattern is robust across all tested threshold definitions. Critically, the collapse scales with dataset complexity: Shapes3D (RGB, 6 factors) shows $1.8\times$ more dead neurons than dSprites (grayscale, 5 factors) under Top-k and $2.2\times$ under L1. These findings establish that interpretability collapse under sparsification is intrinsic to the compression process rather than an artifact of any particular algorithm, training duration, or threshold choice.

cs.LG↗

arXiv2025

HinTel-AlignBench: A Framework and Benchmark for Hindi-Telugu with English-Aligned Samples

With nearly 1.5 billion people and more than 120 major languages, India represents one of the most diverse regions in the world. As multilingual Vision-Language Models (VLMs) gain prominence, robust evaluation methodologies are essential to drive progress toward equitable AI for low-resource languages. Current multilingual VLM evaluations suffer from four major limitations: reliance on unverified auto-translations, narrow task/domain coverage, limited sample sizes, and lack of cultural and natively sourced Question-Answering (QA). To address these gaps, we present a scalable framework to evaluate VLMs in Indian languages and compare it with performance in English. Using the framework, we generate HinTel-AlignBench, a benchmark that draws from diverse sources in Hindi and Telugu with English-aligned samples. Our contributions are threefold: (1) a semi-automated dataset creation framework combining back-translation, filtering, and human verification; (2) the most comprehensive vision-language benchmark for Hindi and and Telugu, including adapted English datasets (VQAv2, RealWorldQA, CLEVR-Math) and native novel Indic datasets (JEE for STEM, VAANI for cultural grounding) with approximately 4,000 QA pairs per language; and (3) a detailed performance analysis of various State-of-the-Art (SOTA) open-weight and closed-source VLMs. We find a regression in performance for tasks in English versus in Indian languages for 4 out of 5 tasks across all the models, with an average regression of 8.3 points in Hindi and 5.5 points for Telugu. We categorize common failure modes to highlight concrete areas of improvement in multilingual multimodal understanding.

cs.CL↗

arXiv2025

GASCADE: Grouped Summarization of Adverse Drug Event for Enhanced Cancer Pharmacovigilance

In the realm of cancer treatment, summarizing adverse drug events (ADEs) reported by patients using prescribed drugs is crucial for enhancing pharmacovigilance practices and improving drug-related decision-making. While the volume and complexity of pharmacovigilance data have increased, existing research in this field has predominantly focused on general diseases rather than specifically addressing cancer. This work introduces the task of grouped summarization of adverse drug events reported by multiple patients using the same drug for cancer treatment. To address the challenge of limited resources in cancer pharmacovigilance, we present the MultiLabeled Cancer Adverse Drug Reaction and Summarization (MCADRS) dataset. This dataset includes pharmacovigilance posts detailing patient concerns regarding drug efficacy and adverse effects, along with extracted labels for drug names, adverse drug events, severity, and adversity of reactions, as well as summaries of ADEs for each drug. Additionally, we propose the Grouping and Abstractive Summarization of Cancer Adverse Drug events (GASCADE) framework, a novel pipeline that combines the information extraction capabilities of Large Language Models (LLMs) with the summarization power of the encoder-decoder T5 model. Our work is the first to apply alignment techniques, including advanced algorithms like Direct Preference Optimization, to encoder-decoder models using synthetic datasets for summarization tasks. Through extensive experiments, we demonstrate the superior performance of GASCADE across various metrics, validated through both automated assessments and human evaluations. This multitasking approach enhances drug-related decision-making and fosters a deeper understanding of patient concerns, paving the way for advancements in personalized and responsive cancer care. The code and dataset used in this work are publicly available.

cs.CL↗

arXiv2025

A Multi-Level Causal Intervention Framework for Mechanistic Interpretability in Variational Autoencoders

Understanding how generative models represent and transform data is a foundational problem in deep learning interpretability. While mechanistic interpretability of discriminative architectures has yielded substantial insights, relatively little work has addressed variational autoencoders (VAEs). This paper presents the first general-purpose multilevel causal intervention framework for mechanistic interpretability of VAEs. The framework comprises four manipulation types: input manipulation, latent-space perturbation, activation patching, and causal mediation analysis. We also define three new quantitative metrics capturing properties not measured by existing disentanglement metrics alone: Causal Effect Strength (CES), intervention specificity, and circuit modularity. We conduct the largest empirical study to date of VAE causal mechanisms across six architectures (standard VAE, beta-VAE, FactorVAE, beta-TC-VAE, DIP-VAE-II, and VQ-VAE) and five benchmarks (dSprites, 3DShapes, MPI3D, CelebA, and SmallNORB), with three seeds per configuration, totaling 90 independent training runs. Our results reveal several findings: (i) a consistent within-dataset negative correlation between CES and DCI disentanglement (the CES-DCI trade-off); (ii) that the KL reweighting mechanism of beta-VAE induces a capacity bottleneck when generative factors approach latent dimensionality, degrading disentanglement on complex datasets; (iii) that no single VAE architecture dominates across all five datasets, with optimal choice depending on dataset structure; and (iv) that CES-based metrics applied to discrete latent spaces (VQ-VAE) yield near-zero values, revealing a critical limitation of continuous-intervention methods for discrete representations. These results provide both a theoretical foundation and comprehensive empirical evaluation for mechanistic interpretability of generative models.

cs.LG↗

arXiv2024

ARDDQN: Attention Recurrent Double Deep Q-Network for UAV Coverage Path Planning and Data Harvesting

Unmanned Aerial Vehicles (UAVs) have gained popularity in data harvesting (DH) and coverage path planning (CPP) to survey a given area efficiently and collect data from aerial perspectives, while data harvesting aims to gather information from various Internet of Things (IoT) sensor devices, coverage path planning guarantees that every location within the designated area is visited with minimal redundancy and maximum efficiency. We propose the ARDDQN (Attention-based Recurrent Double Deep Q Network), which integrates double deep Q-networks (DDQN) with recurrent neural networks (RNNs) and an attention mechanism to generate path coverage choices that maximize data collection from IoT devices and to learn a control scheme for the UAV that generalizes energy restrictions. We employ a structured environment map comprising a compressed global environment map and a local map showing the UAV agent's locate efficiently scaling to large environments. We have compared Long short-term memory (LSTM), Bi-directional long short-term memory (Bi-LSTM), Gated recurrent unit (GRU) and Bidirectional gated recurrent unit (Bi-GRU) as recurrent neural networks (RNN) to the result without RNN We propose integrating the LSTM with the Attention mechanism to the existing DDQN model, which works best on evolution parameters, i.e., data collection, landing, and coverage ratios for the CPP and data harvesting scenarios.

cs.LG↗

arXiv2022

Collisionless Pattern Discovery in Robot Swarms Using Deep Reinforcement Learning

We present a deep reinforcement learning-based framework for automatically discovering patterns available in any given initial configuration of fat robot swarms. In particular, we model the problem of collision-less gathering and mutual visibility in fat robot swarms and discover patterns for solving them using our framework. We show that by shaping reward signals based on certain constraints like mutual visibility and safe proximity, the robots can discover collision-less trajectories leading to well-formed gathering and visibility patterns.

cs.RO↗

arXiv2013

Structural and Electronic Recovery Pathways of a Photoexcited Ultrathin VO2 Film

The structural and electronic recovery pathways of a photoexcited ultrathin VO2 film at nanosecond time scales have been studied using time-resolved x-ray diffraction and transient optical absorption techniques. The recovery pathways from the tetragonal metallic phase to the monoclinic insulating phase are highly dependent on the optical pump fluence. At pump fluences higher than the saturation fluence of 14.7 mJ/cm2, we observed a transient structural state with lattice parameter larger than that of the tetragonal phase, which is decoupled from the metal-to-insulator phase transition. Subsequently, the photoexcited VO2 film recovered to the ground state at characteristic times dependent upon the pump fluence, as a result of heat transport from the film to the substrate. We present a procedure to measure the time-resolved film temperature by correlating photoexcited and temperature-dependent x-ray diffraction measurements. A thermal transport model that incorporates changes of the thermal parameters across the phase transition reproduces the observed recovery dynamics. The optical excitation and fast recovery of ultrathin VO2 films provides a practical method to reversibly switch between the monoclinic insulating and tetragonal metallic state at nanosecond time scales.

cond-mat.str-el↗

arXiv2012

Perpendicular Magnetization and Generic Realization of the Ising Model in Artificial Spin Ice

We have studied frustrated kagome arrays and unfrustrated honeycomb arrays of magnetostatically-interacting single-domain ferromagnetic islands with magnetization normal to the plane. The measured pairwise spin correlations of both lattices can be reproduced by models based solely on nearest-neighbor correlations. The kagome array has qualitatively different magnetostatics but identical lattice topology to previously-studied 'artificial spin ice' systems composed of in-plane moments. The two systems show striking similarities in the development of moment pair correlations, demonstrating a universality in artificial spin ice behavior independent of specific realization in a particular material system.

cond-mat.mes-hall↗

arXiv2012

Magnetization states and switching in narrow-gapped ferromagnetic nanorings

We study permalloy nanorings that are lithographically fabricated with narrow gaps that break the rotational symmetry of the ring while retaining the vortex ground state, using both micromagnetic simulations and magnetic force microscopy (MFM). The vortex chirality in these structures can be readily set with an in-plane magnetic field and easily probed by MFM due to the field associated with the gap, suggesting such rings for possible applications in storage technologies. We find that the gapped ring edge characteristics (i.e., edge profile and gap shape) are critical in determining the magnetization switching field, thus elucidating an essential parameter in the controls of devices that might incorporate such structures.

cond-mat.mes-hall↗

arXiv2011

Programmable Cellular Automata Based Efficient Parallel AES Encryption Algorithm

Cellular Automata(CA) is a discrete computing model which provides simple, flexible and efficient platform for simulating complicated systems and performing complex computation based on the neighborhoods information. CA consists of two components 1) a set of cells and 2) a set of rules . Programmable Cellular Automata(PCA) employs some control signals on a Cellular Automata(CA) structure. Programmable Cellular Automata were successfully applied for simulation of biological systems, physical systems and recently to design parallel and distributed algorithms for solving task density and synchronization problems. In this paper PCA is applied to develop cryptography algorithms. This paper deals with the cryptography for a parallel AES encryption algorithm based on programmable cellular automata. This proposed algorithm based on symmetric key systems.

cs.DC↗