arXiv subjects

Akshay Kulkarni

Publications and source records attributed to Akshay Kulkarni.

At least 19 recordsLinked to original sources

arXiv2026

Entanglement Entropy of Yukawa-Coupled Fields Across a Rindler Horizon

We compute the entanglement entropy across a Rindler horizon in scalar field theory with Yukawa interaction. Starting from a microscopic scalar-mediator theory in flat spacetime, we integrate out the massive mediator to obtain a quadratic but nonlocal effective kernel that determines the ground-state wavefunctional. The reduced density matrix for a single Rindler wedge is constructed explicitly by tracing over the complementary wedge, allowing the entanglement entropy to be evaluated directly from the kernel without replica or geometric methods. Exploiting translational invariance parallel to the horizon, the problem decomposes into independent transverse momentum sectors that reduce effectively to one-dimensional nonlocal systems and can be diagonalized analytically in the weak-coupling regime. The interaction-induced entropy obeys an area law, with leading corrections controlled by the Yukawa screening mass and logarithmically sensitive to the transverse ultraviolet cutoff, reflecting the localization of correlations near the horizon. Although the modular Hamiltonian depends on the Rindler acceleration, the entanglement spectrum and entropy are independent of this choice, demonstrating the observer-independent nature of vacuum entanglement. Our framework provides a direct and microscopically transparent approach to computing interaction-induced corrections to horizon entanglement using nonlocal effective kernels.

hep-th↗

arXiv2026

Entanglement Entropy for Screened Interactions via Dimensional Mapping to Harmonic Oscillators

We investigate interaction-induced corrections to entanglement entropy by mapping a screened Yukawa-type interaction to an effective harmonic oscillator system with controlled anharmonic perturbations. Starting from a one-dimensional interaction $V(x) = -g^2 e^{-\alpha m x}/x$, we reformulate the problem in terms of a four-dimensional radial oscillator, where the finite screening length generates a systematic hierarchy of polynomial interactions in the radial coordinate. This mapping enables a controlled Rayleigh-Schrodinger perturbative treatment of the ground-state wavefunction and an explicit spectral analysis of the reduced density matrix. Working in the weak-screening regime, we compute the leading non-Gaussian correction arising from the quartic interaction $\rho^4$, which appears at order $\alpha^2$ in the expansion of the Yukawa-like potential. We obtain closed analytic expressions for the resulting small eigenvalues of the reduced density matrix and evaluate their contribution to the von Neumann entanglement entropy. We show that the entropy receives analytic corrections at order $\alpha^2$, originating both from explicit anharmonic state-mixing effects and from the implicit $\alpha$ dependence of the Gaussian width parameter. Our results clarify the distinct roles of harmonic renormalization and genuinely non-Gaussian interactions in generating entanglement, establish a systematic power-counting and normalization scheme for higher-order $\rho^{2n}$ perturbations, and provide a transparent oscillator-based framework for computing entanglement entropy in weakly interacting low-dimensional and field-theoretic systems.

quant-ph↗

arXiv2025

Interpretable and Steerable Concept Bottleneck Sparse Autoencoders

Sparse autoencoders (SAEs) promise a unified approach for mechanistic interpretability, concept discovery, and model steering in LLMs and LVLMs. However, realizing this potential requires learned features to be both interpretable and steerable. To that end, we introduce two new computationally inexpensive interpretability and steerability metrics for a systematic analysis of LVLM SAEs. This uncovers two observations; (i) a majority of SAE neurons exhibit either low interpretability or low steerability or both, rendering them ineffective for downstream use; and (ii) user-desired concepts are often absent in the SAE, thus limiting their practical utility. To address these limitations, we propose Concept Bottleneck Sparse Autoencoders (CB-SAE) - a novel post-hoc framework that prunes low-utility neurons and augments the latent space with a lightweight concept bottleneck aligned to a user-defined concept set. The resulting CB-SAE improves interpretability by +32.1% and steerability by +14.5% across LVLMs and image generation tasks.

cs.LG↗

arXiv2025

ReFIne: A Framework for Trustworthy Large Reasoning Models with Reliability, Faithfulness, and Interpretability

Recent advances in long chain-of-thought (CoT) reasoning have largely prioritized answer accuracy and token efficiency, while overlooking aspects critical to trustworthiness. We argue that usable reasoning systems must be trustworthy, characterized by three properties: interpretability, faithfulness, and reliability. To this end, we propose ReFIne, a new training framework that integrates supervised fine-tuning with GRPO to encourage models to: (i) improve interpretability by producing structured, tag-based traces with high-level planning that are easier for humans to follow; (ii) enhance faithfulness by explicitly disclosing the decisive information guiding each solution, with consistent cross-section references; and (iii) promote reliability by providing self-assessments of both the derivation's soundness and the confidence of the final answer. We apply ReFIne to the Qwen3 models at multiple scales (1.7B/4B/8B) and evaluate across mathematical benchmarks of varying difficulty. Our experimental results show that ReFIne models generate clearer and better-structured reasoning traces (interpretability +44.0%), more faithfully expose their underlying decision process (faithfulness +18.8%), and offer informative confidence estimates (reliability +42.4%). These findings highlight an overlooked but important direction: reasoning models should be optimized not only for accuracy, but also for broader dimensions of trustworthiness. Our code is available at: https://github.com/Trustworthy-ML-Lab/Training_Trustworthy_LRM_with_Refine

cs.CL↗

arXiv2025

Design, Fabrication and Characterization of the Thermal Filter Assembly on the Solar Ultraviolet Imaging Telescope (SUIT) on-board Aditya- L1

The Solar Ultraviolet Imaging Telescope (SUIT) observes the Sun in the near-ultraviolet regime on board the Aditya-L1 satellite, India's dedicated mission to study the Sun. SUIT will image the Sun in the wavelength range of 200-400 nm using 11 science bandpasses with varying spectral bandwidths between 0.1-58 nm. Within this range, the Sun provides huge incoming solar flux to the telescope that also varies by a factor of ~ 20 from the lower end to the upper end of the wavelength band of interest. Thermal Filter Assembly (TFA) is an optical component at the SUIT entrance aperture, directly facing the Sun. The TFA is used to control the heat load entering the telescope cavity and also to reduce the signal reaching the SUIT camera system and the charge-coupled device (CCD) sensor, which is limited in full-well capacity and the linear operational regime. The TFA is designed to allow only 0.1-0.45% of the incoming flux to pass within 200-400 nm. The choice of materials for substrate and coating for the filter poses several challenges in terms of contamination, corrosion/ oxidation and durability during the manufacturing process. Additionally, long-term exposure to harsh space environments and the formation of pinholes are other concerns. Direct exposure to the sun leads to a strong temperature gradient along the thickness of the filter. The design and assembly of the TFA are performed to avoid any thermo-elastic stress affecting optical performance. Different levels of qualification tests and the operation of SUIT in orbit for more than 14 months have confirmed the perfect working of the TFA. To the best of our knowledge, the design, development, and testing of such a rejection filter is the first of its kind for space telescopes in the near ultraviolet range.

astro-ph.IM↗

arXiv2025

Beyond Top Activations: Efficient and Reliable Crowdsourced Evaluation of Automated Interpretability

Interpreting individual neurons or directions in activation space is an important topic in mechanistic interpretability. Numerous automated interpretability methods have been proposed to generate such explanations, but it remains unclear how reliable these explanations are, and which methods produce the most accurate descriptions. While crowd-sourced evaluations are commonly used, existing pipelines are noisy, costly, and typically assess only the highest-activating inputs, leading to unreliable results. In this paper, we introduce two techniques to enable cost-effective and accurate crowdsourced evaluation of automated interpretability methods beyond top activating inputs. First, we propose Model-Guided Importance Sampling (MG-IS) to select the most informative inputs to show human raters. In our experiments, we show this reduces the number of inputs needed to reach the same evaluation accuracy by ~13x. Second, we address label noise in crowd-sourced ratings through Bayesian Rating Aggregation (BRAgg), which allows us to reduce the number of ratings per input required to overcome noise by ~3x. Together, these techniques reduce the evaluation cost by ~40x, making large-scale evaluation feasible. Finally, we use our methods to conduct a large scale crowd-sourced study comparing recent automated interpretability methods for vision networks.

cs.CV↗

arXiv2025

Test and Calibration of the Solar Ultraviolet Imaging Telescope (SUIT) on board Aditya-L1

The Solar Ultraviolet Imaging Telescope (SUIT) on board the AdityaL1 mission observes the Sun in the 200-400 nm wavelength range. This paper presents the results of various on ground and on board tests and their comparison with the specifications. Moreover, we also present the scheme for data calibration. We demonstrate that the test results are compliant with the specified figures, except the spatial resolution. Such discrepancy will limit the photometric measurements only, at a scale of 2.2" instead of 1.4" as originally envisioned. The results obtained here show that SUIT observations open up a new window for solar observations.

astro-ph.IM↗

arXiv2025

Interpretable Generative Models through Post-hoc Concept Bottlenecks

Concept bottleneck models (CBM) aim to produce inherently interpretable models that rely on human-understandable concepts for their predictions. However, existing approaches to design interpretable generative models based on CBMs are not yet efficient and scalable, as they require expensive generative model training from scratch as well as real images with labor-intensive concept supervision. To address these challenges, we present two novel and low-cost methods to build interpretable generative models through post-hoc techniques and we name our approaches: concept-bottleneck autoencoder (CB-AE) and concept controller (CC). Our proposed approaches enable efficient and scalable training without the need of real data and require only minimal to no concept supervision. Additionally, our methods generalize across modern generative model families including generative adversarial networks and diffusion models. We demonstrate the superior interpretability and steerability of our methods on numerous standard datasets like CelebA, CelebA-HQ, and CUB with large improvements (average ~25%) over the prior work, while being 4-15x faster to train. Finally, a large-scale user study is performed to validate the interpretability and steerability of our methods.

cs.CV↗

arXiv2025

When Everyday Devices Become Weapons: A Closer Look at the Pager and Walkie-talkie Attacks

Battery-powered technologies like pagers and walkie-talkies have long been integral to civilian and military operations. However, the potential for such everyday devices to be weaponized has largely been underestimated in the realm of cybersecurity. In September 2024, Lebanon experienced a series of unprecedented, coordinated explosions triggered through compromised pagers and walkie-talkies, creating a new category of attack in the domain of cyber-physical warfare. This attack not only disrupted critical communication networks but also resulted in injuries, loss of life, and exposed significant national security vulnerabilities, prompting governments and organizations worldwide to reevaluate their cybersecurity frameworks. This article provides an in-depth investigation into the infamous Pager and Walkie-Talkie attacks, analyzing both technical and non-technical dimensions. Furthermore, the study extends its scope to explore vulnerabilities in other battery-powered infrastructures, such as battery management systems, highlighting their potential exploitation. Existing prevention and detection techniques are reviewed, with an emphasis on their limitations and the challenges they face in addressing emerging threats. Finally, the article discusses emerging methodologies, particularly focusing on the role of physical inspection, as a critical component of future security measures. This research aims to provide actionable insights to bolster the resilience of cyber-physical systems in an increasingly interconnected world.

cs.CR↗

arXiv2025

The Solar Ultraviolet Imaging Telescope on board Aditya-L1

The Solar Ultraviolet Imaging Telescope (SUIT) is an instrument on the Aditya-L1 mission of the Indian Space Research Organization (ISRO) launched on September 02, 2023. SUIT continuously provides, near-simultaneous full-disk and region-of-interest images of the Sun, slicing through the photosphere and chromosphere and covering a field of view up to 1.5 solar radii. For this purpose, SUIT uses 11 filters tuned at different wavelengths in the 200{--}400~nm range, including the Mg~{\sc ii} h~and~k and Ca~{\sc ii}~H spectral lines. The observations made by SUIT help us understand the magnetic coupling of the lower and middle solar atmosphere. In addition, for the first time, it allows the measurements of spatially resolved solar broad-band radiation in the near and mid ultraviolet, which will help constrain the variability of the solar ultraviolet irradiance in a wavelength range that is central for the chemistry of the Earth's atmosphere. This paper discusses the details of the instrument and data products.

astro-ph.SR↗

arXiv2024

Interpretability-Guided Test-Time Adversarial Defense

We propose a novel and low-cost test-time adversarial defense by devising interpretability-guided neuron importance ranking methods to identify neurons important to the output classes. Our method is a training-free approach that can significantly improve the robustness-accuracy tradeoff while incurring minimal computational overhead. While being among the most efficient test-time defenses (4x faster), our method is also robust to a wide range of black-box, white-box, and adaptive attacks that break previous test-time defenses. We demonstrate the efficacy of our method for CIFAR10, CIFAR100, and ImageNet-1k on the standard RobustBench benchmark (with average gains of 2.6%, 4.9%, and 2.8% respectively). We also show improvements (average 1.5%) over the state-of-the-art test-time defenses even under strong adaptive attacks.

cs.CV↗

arXiv2024

Interpreting Neurons in Deep Vision Networks with Language Models

In this paper, we propose Describe-and-Dissect (DnD), a novel method to describe the roles of hidden neurons in vision networks. DnD utilizes recent advancements in multimodal deep learning to produce complex natural language descriptions, without the need for labeled training data or a predefined set of concepts to choose from. Additionally, DnD is training-free, meaning we don't train any new models and can easily leverage more capable general purpose models in the future. We have conducted extensive qualitative and quantitative analysis to show that DnD outperforms prior work by providing higher quality neuron descriptions. Specifically, our method on average provides the highest quality labels and is more than 2$\times$ as likely to be selected as the best explanation for a neuron than the best baseline. Finally, we present a use case providing critical insights into land cover prediction models for sustainability applications. Our code and data are available at https://github.com/Trustworthy-ML-Lab/Describe-and-Dissect.

cs.CV↗

arXiv2023

Aligning Non-Causal Factors for Transformer-Based Source-Free Domain Adaptation

Conventional domain adaptation algorithms aim to achieve better generalization by aligning only the task-discriminative causal factors between a source and target domain. However, we find that retaining the spurious correlation between causal and non-causal factors plays a vital role in bridging the domain gap and improving target adaptation. Therefore, we propose to build a framework that disentangles and supports causal factor alignment by aligning the non-causal factors first. We also investigate and find that the strong shape bias of vision transformers, coupled with its multi-head attention, make it a suitable architecture for realizing our proposed disentanglement. Hence, we propose to build a Causality-enforcing Source-Free Transformer framework (C-SFTrans) to achieve disentanglement via a novel two-stage alignment approach: a) non-causal factor alignment: non-causal factors are aligned using a style classification task which leads to an overall global alignment, b) task-discriminative causal factor alignment: causal factors are aligned via target adaptation. We are the first to investigate the role of vision transformers (ViTs) in a privacy-preserving source-free setting. Our approach achieves state-of-the-art results in several DA benchmarks.

cs.CV↗

arXiv2023

Domain-Specificity Inducing Transformers for Source-Free Domain Adaptation

Conventional Domain Adaptation (DA) methods aim to learn domain-invariant feature representations to improve the target adaptation performance. However, we motivate that domain-specificity is equally important since in-domain trained models hold crucial domain-specific properties that are beneficial for adaptation. Hence, we propose to build a framework that supports disentanglement and learning of domain-specific factors and task-specific factors in a unified model. Motivated by the success of vision transformers in several multi-modal vision problems, we find that queries could be leveraged to extract the domain-specific factors. Hence, we propose a novel Domain-specificity-inducing Transformer (DSiT) framework for disentangling and learning both domain-specific and task-specific factors. To achieve disentanglement, we propose to construct novel Domain-Representative Inputs (DRI) with domain-specific information to train a domain classifier with a novel domain token. We are the first to utilize vision transformers for domain adaptation in a privacy-oriented source-free setting, and our approach achieves state-of-the-art performance on single-source, multi-source, and multi-target benchmarks

cs.CV↗

arXiv2022

Subsidiary Prototype Alignment for Universal Domain Adaptation

Universal Domain Adaptation (UniDA) deals with the problem of knowledge transfer between two datasets with domain-shift as well as category-shift. The goal is to categorize unlabeled target samples, either into one of the "known" categories or into a single "unknown" category. A major problem in UniDA is negative transfer, i.e. misalignment of "known" and "unknown" classes. To this end, we first uncover an intriguing tradeoff between negative-transfer-risk and domain-invariance exhibited at different layers of a deep network. It turns out we can strike a balance between these two metrics at a mid-level layer. Towards designing an effective framework based on this insight, we draw motivation from Bag-of-visual-Words (BoW). Word-prototypes in a BoW-like representation of a mid-level layer would represent lower-level visual primitives that are likely to be unaffected by the category-shift in the high-level features. We develop modifications that encourage learning of word-prototypes followed by word-histogram based classification. Following this, subsidiary prototype-space alignment (SPA) can be seen as a closed-set alignment problem, thereby avoiding negative transfer. We realize this with a novel word-histogram-related pretext task to enable closed-set SPA, operating in conjunction with goal task UniDA. We demonstrate the efficacy of our approach on top of existing UniDA techniques, yielding state-of-the-art performance across three standard UniDA and Open-Set DA object recognition benchmarks.

cs.CV↗

arXiv2022

Concurrent Subsidiary Supervision for Unsupervised Source-Free Domain Adaptation

The prime challenge in unsupervised domain adaptation (DA) is to mitigate the domain shift between the source and target domains. Prior DA works show that pretext tasks could be used to mitigate this domain shift by learning domain invariant representations. However, in practice, we find that most existing pretext tasks are ineffective against other established techniques. Thus, we theoretically analyze how and when a subsidiary pretext task could be leveraged to assist the goal task of a given DA problem and develop objective subsidiary task suitability criteria. Based on this criteria, we devise a novel process of sticker intervention and cast sticker classification as a supervised subsidiary DA problem concurrent to the goal task unsupervised DA. Our approach not only improves goal task adaptation performance, but also facilitates privacy-oriented source-free DA i.e. without concurrent source-target access. Experiments on the standard Office-31, Office-Home, DomainNet, and VisDA benchmarks demonstrate our superiority for both single-source and multi-source source-free DA. Our approach also complements existing non-source-free works, achieving leading performance.

cs.CV↗

arXiv2022

Balancing Discriminability and Transferability for Source-Free Domain Adaptation

Conventional domain adaptation (DA) techniques aim to improve domain transferability by learning domain-invariant representations; while concurrently preserving the task-discriminability knowledge gathered from the labeled source data. However, the requirement of simultaneous access to labeled source and unlabeled target renders them unsuitable for the challenging source-free DA setting. The trivial solution of realizing an effective original to generic domain mapping improves transferability but degrades task discriminability. Upon analyzing the hurdles from both theoretical and empirical standpoints, we derive novel insights to show that a mixup between original and corresponding translated generic samples enhances the discriminability-transferability trade-off while duly respecting the privacy-oriented source-free setting. A simple but effective realization of the proposed insights on top of the existing source-free DA approaches yields state-of-the-art performance with faster convergence. Beyond single-source, we also outperform multi-source prior-arts across both classification and semantic segmentation benchmarks.

cs.CV↗

arXiv2022

Amplitude Spectrum Transformation for Open Compound Domain Adaptive Semantic Segmentation

Open compound domain adaptation (OCDA) has emerged as a practical adaptation setting which considers a single labeled source domain against a compound of multi-modal unlabeled target data in order to generalize better on novel unseen domains. We hypothesize that an improved disentanglement of domain-related and task-related factors of dense intermediate layer features can greatly aid OCDA. Prior-arts attempt this indirectly by employing adversarial domain discriminators on the spatial CNN output. However, we find that latent features derived from the Fourier-based amplitude spectrum of deep CNN features hold a more tractable mapping with domain discrimination. Motivated by this, we propose a novel feature space Amplitude Spectrum Transformation (AST). During adaptation, we employ the AST auto-encoder for two purposes. First, carefully mined source-target instance pairs undergo a simulation of cross-domain feature stylization (AST-Sim) at a particular layer by altering the AST-latent. Second, AST operating at a later layer is tasked to normalize (AST-Norm) the domain content by fixing its latent to a mean prototype. Our simplified adaptation technique is not only clustering-free but also free from complex adversarial alignment. We achieve leading performance against the prior arts on the OCDA scene segmentation benchmarks.

cs.CV↗