Showing 1–50 of 474 results for author: Saha, P

Adaptive Differential Privacy for Federated Medical Image Segmentation Across Diverse Modalities

Authors: Puja Saha, Eranga Ukwatta

Abstract: Large volumes of medical data remain underutilized because centralizing distributed data is often infeasible due to strict privacy regulations and institutional constraints. In addition, models trained in centralized settings frequently fail to generalize across clinical sites because of heterogeneity in imaging protocols and continuously evolving data distributions arising from differences in sca… ▽ More Large volumes of medical data remain underutilized because centralizing distributed data is often infeasible due to strict privacy regulations and institutional constraints. In addition, models trained in centralized settings frequently fail to generalize across clinical sites because of heterogeneity in imaging protocols and continuously evolving data distributions arising from differences in scanners, acquisition parameters, and patient populations. Federated learning offers a promising solution by enabling collaborative model training without sharing raw data. However, incorporating differential privacy into federated learning, while essential for privacy guarantees, often leads to degraded accuracy, unstable convergence, and reduced generalization. In this work, we propose an adaptive differentially private federated learning (ADP-FL) framework for medical image segmentation that dynamically adjusts privacy mechanisms to better balance the privacy-utility trade-off. The proposed approach stabilizes training, significantly improves Dice scores and segmentation boundary quality, and maintains rigorous privacy guarantees. We evaluated ADP-FL across diverse imaging modalities and segmentation tasks, including skin lesion segmentation in dermoscopic images, kidney tumor segmentation in 3D CT scans, and brain tumor segmentation in multi-parametric MRI. Compared with conventional federated learning and standard differentially private federated learning, ADP-FL consistently achieves higher accuracy, improved boundary delineation, faster convergence, and greater training stability, with performance approaching that of non-private federated learning under the same privacy budgets. These results demonstrate the practical viability of ADP-FL for high-performance, privacy-preserving medical image segmentation in real-world federated settings. △ Less

Submitted 7 April, 2026; originally announced April 2026.

Comments: 10 pages, 8 figures. Accepted in SPIE Medical Imaging 2026. Recipient of CAD Best Paper Award: 1st Place, and Robert F. Wagner All-Conference Best Paper Award: Finalist

Journal ref: Proceedings Volume 13926, SPIE Medical Imaging 2026: Computer-Aided Diagnosis

Effects of Spin Fluctuation and Disorder on Topological States of Quasi 2D Ferromagnet Fe1/5CrTe2

Authors: M. Lamba, P. Saha, K. Yadav, N. Kamboj, S. Patnaik

Abstract: We present a thorough magnetization and magneto-transport study of the diluted Fe-intercalated CrTe2 family member, Fe1/5CrTe2, a van der Waals ferromagnet. Fe1/5CrTe2 shows an elevated Curie transition temperature of 182 K in comparison to the Fe1/3CrTe2 composition, indicating the sensitive role of Fe concentration in modulating magnetic exchange interactions within the CrTe2 framework. The satu… ▽ More We present a thorough magnetization and magneto-transport study of the diluted Fe-intercalated CrTe2 family member, Fe1/5CrTe2, a van der Waals ferromagnet. Fe1/5CrTe2 shows an elevated Curie transition temperature of 182 K in comparison to the Fe1/3CrTe2 composition, indicating the sensitive role of Fe concentration in modulating magnetic exchange interactions within the CrTe2 framework. The saturated magnetization exhibits a quadratic dependence with temperature, indicating the presence of long-wavelength spin fluctuations. Analysis of the temperature dependent resistivity reveals a dominant T3/2 contribution over the typical T2 behavior, signaling substantial coupling between conduction electrons and localized spins. The magnetoresistance shows a linear and non-saturating negative field dependency throughout a wide temperature range below TC, which is compatible with the increasing suppression of spin-disorder dispersion related to ferromagnetic spin fluctuations. A thorough analysis of the anomalous Hall effect (AHE) shows that extrinsic skew-scattering contribution, which is associated to Fe-related disorder, dominates the anomalous Hall response. The systematic separation of intrinsic and extrinsic components reveals that, over a wide temperature range, the intrinsic anomalous Hall conductivity scales linearly with the saturation magnetization, despite the substantial extrinsic dominant background. The linear behavior of intrinsic anomalous Hall conductivity with magnetization is in line with a long wavelength spin-fluctuation framework, where thermal spin disorder lowers net magnetization without significantly altering the underlying electronic structure. These findings reveal Fe1/5CrTe2 as a newly investigated van der Waals ferromagnet where spin fluctuations and disorder coexist with a well-defined intrinsic Berry-curvature contribution to the Hall response. △ Less

Submitted 6 April, 2026; originally announced April 2026.

Nonlinear Lattice Framework for Inflation: Bridging stochastic inflation and the $δ{N}$ formalism

Authors: Pankaj Saha, Yuichiro Tada, Yuko Urakawa

Abstract: Understanding when inflationary perturbations become genuinely nonlinear near the horizon crossing requires methods that go beyond both linear perturbation theory and the gradient expansion. In this work, we introduce a nonlinear lattice framework for single-field inflation based on a shear-free, locally Friedmann-Lemaître-Robertson-Walker geometry. This approach captures inhomogeneous local expan… ▽ More Understanding when inflationary perturbations become genuinely nonlinear near the horizon crossing requires methods that go beyond both linear perturbation theory and the gradient expansion. In this work, we introduce a nonlinear lattice framework for single-field inflation based on a shear-free, locally Friedmann-Lemaître-Robertson-Walker geometry. This approach captures inhomogeneous local expansion rates, curvature contributions to the local Friedmann equation, and proper-volume weighting at a fraction of the computational cost of full numerical relativity. We construct fully nonlinear $δN$ observables on uniform-density slices, together with other practical time-dependent estimators for the curvature perturbations. After validating the framework in a standard slow-roll regime, we apply it to Starobinsky's linear-potential model featuring an intermittent ultra-slow-roll (USR) phase and a sharp potential transition. During this non-attractor USR regime, the lattice captures the separation of curvature perturbation estimators, the growth and subsequent stabilisation of non-Gaussianity, and a transient weakening of the shear-free approximation when the inflaton velocity becomes very small. Our framework provides a practical intermediate approach between rigid background lattice simulations and full numerical relativity, offering a nonlinear bridge between lattice methods, the $δN$ formalism, and the stochastic inflation formalism. △ Less

Submitted 10 April, 2026; v1 submitted 1 April, 2026; originally announced April 2026.

Comments: Corrected a figure, added references, and made minor revisions. 39 Pages, 8 figures

Report number: KEK-TH-2821, KEK-Cosmo-0415, RUP-26-4

Gigahertz-clocked Generation of Highly Indistinguishable Photons at C-band Wavelengths

Authors: Robert Behrends, Lucas Rickert, Nils D. Kewitz, Martin v. Helversen, Pratim K. Saha, Mareike Lach, Jochen Kaupp, Yorick Reum, Tobias-Huber-Loyola, Sven Höfling, Andreas Pfenning, Tobias Heindel

Abstract: High-performance single-photon sources at telecom C-band wavelentghs are key building blocks for applications in long-distance quantum communication. Here, we report the generation of highly indistinguishable, single photons at a clock-rate of 2.5 GHz. This is achieved by coherently driving the biexciton transition ($T_1^\mathrm{XX}=64(1)\,$ps) of a semiconductor quantum dot embedded in a microcav… ▽ More High-performance single-photon sources at telecom C-band wavelentghs are key building blocks for applications in long-distance quantum communication. Here, we report the generation of highly indistinguishable, single photons at a clock-rate of 2.5 GHz. This is achieved by coherently driving the biexciton transition ($T_1^\mathrm{XX}=64(1)\,$ps) of a semiconductor quantum dot embedded in a microcavity with strong asymmetric Purcell enhancement. Employing pulsed two-photon resonant excitation, strong multiphoton suppression with $g^{(2)}(0) < 4\%$ and high two-photon-interference visibility of $V_\mathrm{raw}> 85\%$ is observed. The observed photon indistinguishability is close to the theoretical limit expected for the photonically engineered radiative cascade and matches values obtained at lower repetition rates. Our results show a substantial advancement towards interference-based quantum information protocols at unprecedented data rates in the telecom C-Band. △ Less

Submitted 1 April, 2026; v1 submitted 27 March, 2026; originally announced March 2026.

Comments: 6 pages, 2 figures

Estimation of the magnetic field strength from ALMA dust polarization in the protocluster G327.29

Authors: A. Koley, P. Sanhueza, A. M. Stutz, P. Saha, F. A. Olguin, A. Ginsburg, N. Sandoval-Garrido, N. Castro-Toledo

Abstract: Magnetic fields and turbulence may play a crucial role in the evolution of molecular clouds and ultimately in the formation of dense cores and stars. Despite being studied in many molecular clouds, the exact role of magnetic fields and turbulence in star formation is still poorly understood. Here, we report the high resolution plane of sky magnetic field (B_pos) morphology toward the high mass sta… ▽ More Magnetic fields and turbulence may play a crucial role in the evolution of molecular clouds and ultimately in the formation of dense cores and stars. Despite being studied in many molecular clouds, the exact role of magnetic fields and turbulence in star formation is still poorly understood. Here, we report the high resolution plane of sky magnetic field (B_pos) morphology toward the high mass star forming region G327.29, obtained with the 12-meter of the Atacama Large Millimeter/sub-millimeter Array (ALMA) telescope. From our analysis, we obtain a complex B_pos morphology where the magnetic field orientation is uniformly distributed across the entire range from -90 to +90 deg. The observed area is composed of one filament and one dense central clump, which harbor multiple dense cores. The total magnetic field strengths (B_tot) in these regions are 1.4 \pm 0.7 mG and 2.0 \pm 0.8 mG at a number density (n) of 6.8 \pm 1.5 x 10^5 and 1.1 \pm 0.3 x 10^6 cm^-3 , derived from the angular dispersion function (ADF) method. The virial parameters (α vir )in these regions are 7.7 \pm 7.1 and 0.7 \pm 0.6, suggesting that the regions may be gravitationally bound or unbound after accounting for the errors. Moreover, the ratio of turbulent to magnetic energy (~ 0.25) indicates that the magnetic field is dynamically more important than turbulence. The relative influence of turbulence and magnetic fields on core dynamics appears to depend on how the B_tot scales with gas density (\r{ho}) in the densest regions. In summary, this work presents a comprehensive analysis of the relative roles of magnetic fields, turbulence, and gravity in regulating high-mass star formation in G327.29, enabled by high-resolution ALMA observations. △ Less

Submitted 24 March, 2026; originally announced March 2026.

Comments: Submitted to A&A journal

When Data Protection Fails to Protect: Law, Power, and Postcolonial Governance in Bangladesh

Authors: Pratyasha Saha, Anita Say Chan, Sharifa Sultana

Abstract: Rapid digitization across government services, financial platforms, and telecommunications has intensified the collection and processing of large scale personal data in Bangladesh. In response, the state has introduced multiple regulatory instruments, including the Personal Data Protection Ordinance, the Cyber Security Ordinance, and the National Data Governance Ordinance in 2025. While these init… ▽ More Rapid digitization across government services, financial platforms, and telecommunications has intensified the collection and processing of large scale personal data in Bangladesh. In response, the state has introduced multiple regulatory instruments, including the Personal Data Protection Ordinance, the Cyber Security Ordinance, and the National Data Governance Ordinance in 2025. While these initiatives signal an emerging legal regime for data protection, little scholarly work examines how these frameworks operate collectively in practice. This paper presents a legal and institutional analysis of Bangladeshs emerging data protection regime through a systematic review of these three ordinances. Through this review, the paper provides an integrated mapping of Bangladeshs evolving data protection framework and identifies key legal and institutional barriers that undermine the effective protection of citizens personal data. Our findings reveal that this emerging regime is constrained by limited institutional independence, uneven regulatory capacity, and the misaligned legal assumption of individualized, autonomous data subjects. Furthermore, these frameworks invisibilize prevalent sociotechnical layers, such as informal data flows and mediated access via human bridges, rendering formal protections difficult to operationalize. This paper contributes to HCI scholarship by expanding the concept of data protection as a complex sociotechnical design problem shaped by the informal infrastructures of the Global South. △ Less

Submitted 23 March, 2026; originally announced March 2026.

Mixture of Chapters: Scaling Learnt Memory in Transformers

Authors: Tasmay Pankaj Tibrewal, Pritish Saha, Ankit Meda, Kunal Singh, Pradeep Moturi

Abstract: Transformers lack an explicit architectural mechanism for storing and organizing knowledge acquired during training. We introduce learnable sparse memory banks: a set of latent tokens, randomly initialized and trained end-to-end, that transformer layers query via cross-attention to retrieve stored knowledge. To scale memory capacity without prohibitive attention costs, we propose chapter-based rou… ▽ More Transformers lack an explicit architectural mechanism for storing and organizing knowledge acquired during training. We introduce learnable sparse memory banks: a set of latent tokens, randomly initialized and trained end-to-end, that transformer layers query via cross-attention to retrieve stored knowledge. To scale memory capacity without prohibitive attention costs, we propose chapter-based routing inspired by Mixture-of-Experts architectures, partitioning the memory bank into chapters and training a router to select relevant subsets per input. This enables scaling to 262K memory tokens while maintaining tractable computation. We evaluate our approach against standard transformers (in iso-FLOP settings) on pre-training and instruction fine-tuning across relevant benchmarks. Our models surpass iso-FLOP baselines suggesting scope for a new axis of scaling, demonstrating that explicit associative memory provides complementary capacity to what is captured implicitly in model parameters. Additionally, we observe improved knowledge retention under continued training, with robustness to forgetting when transitioning between training phases (e.g., pretraining to instruction fine-tuning). △ Less

Submitted 22 March, 2026; originally announced March 2026.

Comments: 20 pages, 2 figures, 8 tables. Accepted at ICLR 2026 New Frontiers in Associative Memory Workshop. Code available at https://github.com/Tasmay-Tibrewal/Memory

When turbulence beats magnetism: origin of massive star cluster seeds

Authors: Junhao Liu, Patricio Sanhueza, Piyali Saha, Kaho Morii, Josep Miquel Girart, Qizhou Zhang, Fumitaka Nakamura, Paulo C. Cortes, Valeska Valdivia, Benoit Commercon, Patrick M. Koch, Kate Pattle, Xing Lu, Janik Karoly, Manuel Fernandez-Lopez, Ian W. Stephens, Huei-Ru Vivien Chen, Chi-Yan Law, Keping Qiu, Shanghuo Li, Henrik Beuther, Eun Jung Chung, Jia-Wei Wang, Fernando A. Olguin, Yu Cheng , et al. (10 additional authors not shown)

Abstract: High-mass stars form in protoclusters, where gravo-magnetic processes shape collapsing clouds and clumps to be elongated preferentially perpendicular to magnetic (B) fields. Yet it remains unclear whether gravo-magnetic processes still govern the formation of smaller-scale condensations in massive-star-forming protoclusters, which are crucial for understanding the stellar initial mass function and… ▽ More High-mass stars form in protoclusters, where gravo-magnetic processes shape collapsing clouds and clumps to be elongated preferentially perpendicular to magnetic (B) fields. Yet it remains unclear whether gravo-magnetic processes still govern the formation of smaller-scale condensations in massive-star-forming protoclusters, which are crucial for understanding the stellar initial mass function and multiplicity. Here we report the first statistical evidence that the condensation elongations are preferentially aligned with local B fields, based on high-resolution data from the largest dust polarization survey toward 30 massive star-forming regions with the Atacama Large Millimeter/submillimeter Array (ALMA). Our clustered massive star formation simulations reveal that this more parallel alignment is exclusively observed in models where initial turbulence dominates B fields. In contrast, models with initial B fields dominating turbulence distinctly exhibit a more perpendicular alignment. The comparison between observations and simulations suggests that turbulence could play a more important role than B fields in the formation of condensations in the context of clustered massive star formation, contradicting the prediction of classical magnetically regulated models. Moreover, we find a possibly turbulence-induced preferential misalignment between the B field and rotation axis of condensations, which may potentially reduce the magnetic braking efficiency and facilitate the formation of large protostellar disks. △ Less

Submitted 17 March, 2026; originally announced March 2026.

Comments: 30 pages, 13 figures. Initially submitted to Nature Astronomy on October 27, 2025. The uploaded arXiv version is before the final revision

Can Safety Emerge from Weak Supervision? A Systematic Analysis of Small Language Models

Authors: Punyajoy Saha, Sudipta Halder, Debjyoti Mondal, Subhadarshi Panda

Abstract: Safety alignment is critical for deploying large language models (LLMs) in real-world applications, yet most existing approaches rely on large human-annotated datasets and static red-teaming benchmarks that are costly, difficult to scale, and slow to adapt to evolving model behaviors. Moreover, overly conservative safety mechanisms can reduce model usefulness by rejecting sensitive but legitimate… ▽ More Safety alignment is critical for deploying large language models (LLMs) in real-world applications, yet most existing approaches rely on large human-annotated datasets and static red-teaming benchmarks that are costly, difficult to scale, and slow to adapt to evolving model behaviors. Moreover, overly conservative safety mechanisms can reduce model usefulness by rejecting sensitive but legitimate queries. We introduce Self-MOA (Self Multi-Objective Alignment), a fully automated framework for aligning small language models using weak supervision from automated evaluator models. Self-MOA operates as a closed loop that dynamically generates model-specific red team prompts, constructs preference data from model-generated responses, and aligns models via multi-objective preference optimization to jointly optimize for safety and helpfulness. Across multiple small language models and safety benchmarks, Self-MOA achieves a 12.41\% improvement in safety while preserving helpfulness, using as little as 11 times less training data than human-supervised alignment baselines. These results demonstrate that adaptive, automated alignment can reduce the dependence on static, human-curated safety pipelines in resource-constrained settings. △ Less

Submitted 6 March, 2026; originally announced March 2026.

Comments: 19 pages, 10 tables, 7 figures, under Review

Experience-Guided Self-Adaptive Cascaded Agents for Breast Cancer Screening and Diagnosis with Reduced Biopsy Referrals

Authors: Pramit Saha, Mohammad Alsharid, Joshua Strong, J. Alison Noble

Abstract: We propose an experience-guided cascaded multi-agent framework for Breast Ultrasound Screening and Diagnosis, called BUSD-Agent, that aims to reduce diagnostic escalation and unnecessary biopsy referrals. Our framework models screening and diagnosis as a two-stage, selective decision-making process. A lightweight `screening clinic' agent, restricted to classification models as tools, selectively f… ▽ More We propose an experience-guided cascaded multi-agent framework for Breast Ultrasound Screening and Diagnosis, called BUSD-Agent, that aims to reduce diagnostic escalation and unnecessary biopsy referrals. Our framework models screening and diagnosis as a two-stage, selective decision-making process. A lightweight `screening clinic' agent, restricted to classification models as tools, selectively filters out benign and normal cases from further diagnostic escalation when malignancy risk and uncertainty are estimated as low. Cases that have higher risks are escalated to the `diagnostic clinic' agent, which integrates richer perception and radiological description tools to make a secondary decision on biopsy referral. To improve agent performance, past records of pathology-confirmed outcomes along with image embeddings, model predictions, and historical agent actions are stored in a memory bank as structured decision trajectories. For each new case, BUSD-Agent retrieves similar past cases based on image, model response and confidence similarity to condition the agent's current decision policy. This enables retrieval-conditioned in-context adaptation that dynamically adjusts model trust and escalation thresholds from prior experiences without parameter updates. Evaluation across 10 breast ultrasound datasets shows that the proposed experience-guided workflow reduces diagnostic escalation in BUSD-Agent from 84.95% to 58.72% and overall biopsy referrals from 59.50% to 37.08%, compared to the same architecture without trajectory conditioning, while improving average screening specificity by 68.48% and diagnostic specificity by 6.33%. △ Less

Submitted 27 February, 2026; originally announced February 2026.

PRODIGE - envelope to disk with NOEMA: VII. (Complex) organic molecules in the NGC1333 IRAS4B1 outflow: A new laboratory for shock chemistry

Authors: Laura A. Busch, J. E. Pineda, P. Caselli, D. M. Segura-Cox, S. Narayanan, C. Gieser, M. J. Maureira, T. -H. Hsieh, Y. Lin, M. T. Valdivia-Mena, L. Bouscasse, Th. Henning, D. Semenov, A. Fuente, Y. -R. Chou, L. Mason, P. C. Cortés, L. W. Looney, I. W. Stephens, M. Tafalla, A. Dutrey, W. Kwon, P. Saha

Abstract: Shock chemistry is an excellent tool to shed light on the formation and destruction mechanisms of complex organic molecules (COMs). The L1157-mm outflow is the only low-mass protostellar outflow that has extensively been studied in this regard. Using the data taken as part of the PRODIGE (PROtostars & DIsks: Global Evolution) large program, we aim to map COM emission and derive the molecular compo… ▽ More Shock chemistry is an excellent tool to shed light on the formation and destruction mechanisms of complex organic molecules (COMs). The L1157-mm outflow is the only low-mass protostellar outflow that has extensively been studied in this regard. Using the data taken as part of the PRODIGE (PROtostars & DIsks: Global Evolution) large program, we aim to map COM emission and derive the molecular composition of the protostellar outflow driven by the Class 0 protostar NGC1333 IRAS4B1 to introduce it as a new laboratory to study the impact of shocks on COM chemistry. In addition to typical outflow tracers such as SiO and CO, outflow emission is seen from H2CO, HNCO, and HC3N, as well as from the COMs CH3OH, CH3CN, and CH3CHO, and even from deuterated species such as DCN, D2CO, and CH2DOH. Maps of integrated intensity ratios between CH3OH and DCN, D2CO, and CH3CHO reveal gradients with distance from the protostar. Intensity ratio maps of HC3N and CH3CN with respect to CH3OH peak in the southern lobe where temperatures are highest. Rotational temperatures derived towards two positions, one in each lobe, are found in the range ~50-100 K. Abundances with respect to CH3OH are higher by factors of a few than for the L1157-B1. In conclusion, for the first time, we securely detected the COMs CH3CN, CH3CHO, and CH2DOH in the IRAS 4B1 outflow, serendipitously with limited sensitivity and bandwidth. Targeted observations will enable the discovery of new COMs and a more detailed analysis of their emission. Morphological differences between molecules in the IRAS 4B1 outflow lobes and their relative abundances provide first proof that this outflow is a promising new laboratory for shock chemistry, which will offer crucial information on COM formation and destruction as well as outflow structure and kinematics. △ Less

Submitted 19 February, 2026; originally announced February 2026.

Comments: Accepted for publication in A&A

Picking the Right Specialist: Attentive Neural Process-based Selection of Task-Specialized Models as Tools for Agentic Healthcare Systems

Authors: Pramit Saha, Joshua Strong, Mohammad Alsharid, Divyanshu Mishra, J. Alison Noble

Abstract: Task-specialized models form the backbone of agentic healthcare systems, enabling the agents to answer clinical queries across tasks such as disease diagnosis, localization, and report generation. Yet, for a given task, a single "best" model rarely exists. In practice, each task is better served by multiple competing specialist models where different models excel on different data samples. As a re… ▽ More Task-specialized models form the backbone of agentic healthcare systems, enabling the agents to answer clinical queries across tasks such as disease diagnosis, localization, and report generation. Yet, for a given task, a single "best" model rarely exists. In practice, each task is better served by multiple competing specialist models where different models excel on different data samples. As a result, for any given query, agents must reliably select the right specialist model from a heterogeneous pool of tool candidates. To this end, we introduce ToolSelect, which adaptively learns model selection for tools by minimizing a population risk over sampled specialist tool candidates using a consistent surrogate of the task-conditional selection loss. Concretely, we propose an Attentive Neural Process-based selector conditioned on the query and per-model behavioral summaries to choose among the specialist models. Motivated by the absence of any established testbed, we, for the first time, introduce an agentic Chest X-ray environment equipped with a diverse suite of task-specialized models (17 disease detection, 19 report generation, 6 visual grounding, and 13 VQA) and develop ToolSelectBench, a benchmark of 1448 queries. Our results demonstrate that ToolSelect consistently outperforms 10 SOTA methods across four different task families. △ Less

Submitted 16 February, 2026; originally announced February 2026.

Observations of Binary Stars with the 1.3-m Devasthal Fast Optical Telescope Using Speckle Interferometry: An Attempt

Authors: Km Nitu Rai, Arjun Dawn, Neelam Panwar, Jeewan C Pandey, Subrata Sarangi, Prasenjit Saha

Abstract: We present a feasibility study exploring the implementation of optical interferometry and speckle techniques with the 1.3-m Devasthal Fast Optical Telescope (DFOT) at ARIES, which is currently dedicated to photometric observations. Using the sCMOS camera as the DFOT backend, we perform interferometric speckle observations of several binary stars. Standard Speckle Interferometry (SI) algorithms are… ▽ More We present a feasibility study exploring the implementation of optical interferometry and speckle techniques with the 1.3-m Devasthal Fast Optical Telescope (DFOT) at ARIES, which is currently dedicated to photometric observations. Using the sCMOS camera as the DFOT backend, we perform interferometric speckle observations of several binary stars. Standard Speckle Interferometry (SI) algorithms are applied to analyze the recorded data. While this study does not aim to achieve the diffraction limit of DFOT or address a full science-driven resolution case, it serves as a crucial testbed for instrumentation, data acquisition, and analysis of Speckles with DFOT. Notably, we successfully identify and correct tracking-related positional errors in the observed binary systems, demonstrating the viability of the approach. These results provide strong motivation for more systematic observations and future implementation of optical interferometry techniques at meter-class telescopes. △ Less

Submitted 13 February, 2026; originally announced February 2026.

Comments: Submitted to BINA Conference-Proceedings. Comments are welcome

ESO White Paper on Intensity Interferometry: Cosmology, Fundamental Physics, Quantum Optics

Authors: Robin Kaiser, William Guerin, Farrokh Vakili, Jean-Philippe Berger, Andrei Nomerotski, Sergei Kulkov, Peter Svihra, Eva Santos, Colin Carlile, Dainis Dravins, Stefan Funk, Prasenjit Saha, Roland Walter, Marcelo Borges Fernandes, Alex G. Kim, David Dunsky, Ken Van Tilburg, Masha Baryakhtar, Marios Galanis, Robert V. Wagoner, Neal Dalal, Junwu Huang, Charles Gammie, Norman W. Murray

Abstract: In this whitepaper, we outline how recent technological advances and ongoing developments open qualitatively new science opportunities in cosmology, fundamental physics, and quantum astrophysics. First, intensity interferometry can contribute to one of the most foundational observables in cosmology: the expansion rate of the Universe. Its angular resolution allows it to resolve the angular extent… ▽ More In this whitepaper, we outline how recent technological advances and ongoing developments open qualitatively new science opportunities in cosmology, fundamental physics, and quantum astrophysics. First, intensity interferometry can contribute to one of the most foundational observables in cosmology: the expansion rate of the Universe. Its angular resolution allows it to resolve the angular extent of extragalactic objects such as supernovae or quasars; combined with a physical scale local to the source, this yields an angular diameter distance and hence a 'Hubble diagram'. Second, the nature of dark matter can be probed via the astrometric lensing signatures of tiny dark matter halos. Third, intensity interferometry gives direct access to second-order coherence properties of astrophysical emission, opening a window onto genuinely quantum aspects of astrophysical light. △ Less

Submitted 13 February, 2026; originally announced February 2026.

Comments: ESO White Paper on Intensity Interferometry 4 pages

Highly-Indistinguishable Single-Photons at 1550 nm from a Two-photon Resonantly Excited Purcell-enhanced Quantum Dot

Authors: Robert Behrends, Martin v. Helversen, Pratim K. Saha, Lucas Rickert, Koray Kaymazlar, Mareike Lach, Nils D. Kewitz, Jochen Kaupp, Yorick Reum, Tobias Huber-Loyola, Sven Höfling, Andreas Pfenning, Tobias Heindel

Abstract: In this work we present a cavity-enhanced InAs/$\mathrm{In_{0.53}Al_{0.23}Ga_{0.24}As}$ quantum dot (QD) single-photon source in the telecom C-band with a record-low biexciton emitter decay time of \SI{67.4(2)}{ps} under resonant two-photon excitation (TPE). We observe strong multiphoton suppression associated with $g^{(2)}_\mathrm{X}(0) = 0.006(1)$ and $g^{(2)}_\mathrm{XX}(0) = 0.007(1)$ for the… ▽ More In this work we present a cavity-enhanced InAs/$\mathrm{In_{0.53}Al_{0.23}Ga_{0.24}As}$ quantum dot (QD) single-photon source in the telecom C-band with a record-low biexciton emitter decay time of \SI{67.4(2)}{ps} under resonant two-photon excitation (TPE). We observe strong multiphoton suppression associated with $g^{(2)}_\mathrm{X}(0) = 0.006(1)$ and $g^{(2)}_\mathrm{XX}(0) = 0.007(1)$ for the exciton (X) and biexciton (XX) emission, respectively. Due to a asymmetric Purcell enhancement of the XX-X cascade, the two-photon interference (TPI) visibility of XX photons under $π$-pulse excitation of $V_{\rm{TPI}} = 90(3)\%$ reaches the theoretical limit and clearly exceeds the $\sim60\%$ expected for standard XX-X cascades without photonic engineering. Furthermore, adding a second timed laser pulse coinciding with XX emission energy, we demonstrate stimulated TPE in the telecom C-Band. The result is an improved TPI visibility of the X photons of $V_{\rm{TPI}}=0.69(3)$ compared to TPE with $V_{\rm{TPI}}=0.61(4)$, with both being reduced compared to the theoretical values due to present dephasing effects. The advances presented in this work hold important promises for the implementation of advanced schemes of quantum communication using deterministic quantum light sources. △ Less

Submitted 5 February, 2026; originally announced February 2026.

Comments: 9 pages, 3 figures

Clustering-driven Memory Compression for On-device Large Language Models

Authors: Ondrej Bohdal, Pramit Saha, Umberto Michieli, Mete Ozay, Taha Ceritli

Abstract: Large language models (LLMs) often rely on user-specific memories distilled from past interactions to enable personalized generation. A common practice is to concatenate these memories with the input prompt, but this approach quickly exhausts the limited context available in on-device LLMs. Compressing memories by averaging can mitigate context growth, yet it frequently harms performance due to se… ▽ More Large language models (LLMs) often rely on user-specific memories distilled from past interactions to enable personalized generation. A common practice is to concatenate these memories with the input prompt, but this approach quickly exhausts the limited context available in on-device LLMs. Compressing memories by averaging can mitigate context growth, yet it frequently harms performance due to semantic conflicts across heterogeneous memories. In this work, we introduce a clustering-based memory compression strategy that balances context efficiency and personalization quality. Our method groups memories by similarity and merges them within clusters prior to concatenation, thereby preserving coherence while reducing redundancy. Experiments demonstrate that our approach substantially lowers the number of memory tokens while outperforming baseline strategies such as naive averaging or direct concatenation. Furthermore, for a fixed context budget, clustering-driven merging yields more compact memory representations and consistently enhances generation quality. △ Less

Submitted 24 January, 2026; originally announced January 2026.

Comments: Accepted at ICASSP 2026

Polynomial Chaos-based Input Shaper Design under Time-Varying Uncertainty

Authors: Johannes Güttler, Karan Baker, Premjit Saha, James Warner, Adrian Stein

Abstract: The work presented here investigates the application of polynomial chaos expansion toward input shaper design in order to maintain robustness in dynamical systems subject to uncertainty. Furthermore, this work intends to specifically address time-varying uncertainty by employing intrusive polynomial chaos expansion. The methodology presented is validated through numerical simulation of intrusive p… ▽ More The work presented here investigates the application of polynomial chaos expansion toward input shaper design in order to maintain robustness in dynamical systems subject to uncertainty. Furthermore, this work intends to specifically address time-varying uncertainty by employing intrusive polynomial chaos expansion. The methodology presented is validated through numerical simulation of intrusive polynomial chaos expansion formulation applied to spring mass system experiencing time-varying uncertainty in the spring stiffness. The system also evaluates non-robust and robust input shapers through the framework in order to identify designs that minimize residual energy. Results indicate that vibration mitigation is achieved at a similar accuracy, yet at higher efficiency compared to a Monte Carlo framework. △ Less

Submitted 3 February, 2026; v1 submitted 23 January, 2026; originally announced January 2026.

Five-point partial waves, splitting constraints and hidden zeros

Authors: Arnab Priya Saha, Aninda Sinha

Abstract: We study the partial-wave expansion of residues of five-point tree-amplitude involving identical scalar particles in the external legs. We check the construction using massive spinor-helicity building blocks and by matching to the tree-level five-point Veneziano amplitude at fixed mass levels. As an application, we express five-point splitting constraints - the reduction of the five-point amplitud… ▽ More We study the partial-wave expansion of residues of five-point tree-amplitude involving identical scalar particles in the external legs. We check the construction using massive spinor-helicity building blocks and by matching to the tree-level five-point Veneziano amplitude at fixed mass levels. As an application, we express five-point splitting constraints - the reduction of the five-point amplitude to products of four-point amplitudes on special kinematic loci - as linear relations among the five-point partial-wave coefficients. At low mass levels these constraints, together with spin truncation, fix the full five-point partial-wave data in terms of the four-point coefficients and imply simple compatibility conditions; remarkably, imposing two independent splitting loci also forces the residue to vanish on their intersection, making the associated hidden zero manifest in partial-wave space. We also show that once both channels allow spin-2 exchange a genuine kernel can remain, indicating the need for additional higher-point input to achieve complete rigidity. △ Less

Submitted 21 January, 2026; originally announced January 2026.

Comments: 41 pages, 4 figures

Magnetic field morphological diagnostics with ALMA in the G327.29 protocluster: VGT versus dust polarization

Authors: A. Koley, A. M. Stutz, A. Lazarian, Y. Hu, P. Sanhueza, P. Saha, R. H. Alvarez-Gutierrez, N. S. Sandoval-Garrido, N. Castro-Toledo, G. Bernal Mesina

Abstract: Magnetic fields and turbulence may play a key role in the evolution of protoclusters, influencing the formation of dense cores and stars. Here, we examine the morphology of the magnetic fields in the G327.29 protocluster using both the velocity gradient technique (VGT) extracted from molecular line emissions and linear polarization in the dust continuum emission. The VGT analysis is performed usin… ▽ More Magnetic fields and turbulence may play a key role in the evolution of protoclusters, influencing the formation of dense cores and stars. Here, we examine the morphology of the magnetic fields in the G327.29 protocluster using both the velocity gradient technique (VGT) extracted from molecular line emissions and linear polarization in the dust continuum emission. The VGT analysis is performed using four molecular tracers: DCN (3-2), C18O (2-1), HN13C (3-2), and H13CO+ (3-2) - which probe gas across different density regimes, observed with the ALMA 12 m array. Owing to its sensitivity to gas dynamics, a comparison between VGT and dust polarization provides a powerful probe of the evolutionary processes in massive star-forming regions. From our analysis we reveal a complex magnetic-field structure, shaped by the combined influence of turbulence and gravity. In addition, it also appears that there is a large-scale (beyond the core scale) gravitational infall from the surrounding medium on to the filament and the central densest region. Furthermore, we observe that cores are dominated by a mix of turbulence and gravity. Overall, this work presents, likely for the first time, the application of VGT to a massive protocluster, G327.29, using high-resolution ALMA observations. △ Less

Submitted 19 January, 2026; originally announced January 2026.

Comments: Submitted to Astronomy & Astrophysics (A&A) journal

Perception of Deepfakes among Bangladeshi Women

Authors: Sharifa Sultana, Pratyasha Saha, Nadira Nowsher, Sumaia Arefin Ritu, Zinnat Sultana, Syed Ishtiaque Ahmed, S M Taiabul Haque

Abstract: As deepfake technology becomes more accessible, concerns about its misuse and societal impact are escalating, particularly in regions like the Global South where digital literacy and regulatory measures are often limited. While previous research has explored deepfakes in contexts such as detection and media manipulation, there is a noticeable gap in understanding how individuals in these regions p… ▽ More As deepfake technology becomes more accessible, concerns about its misuse and societal impact are escalating, particularly in regions like the Global South where digital literacy and regulatory measures are often limited. While previous research has explored deepfakes in contexts such as detection and media manipulation, there is a noticeable gap in understanding how individuals in these regions perceive and interact with deepfake media. This study addresses this gap by investigating how Bangladeshi women perceive deepfakes and the socio-cultural factors influencing their awareness, concerns, and responses to this technology. Drawing on 15 semi-structured interviews, we uncover how cultural values, gendered norms, trust in institutions, and the prevalence of digital harassment shape their perceptions and coping mechanisms. Through this research, we aim to advance existing scholarship in HCI by offering insights into the design of culturally sensitive interventions, educational initiatives, and policy frameworks to address the challenges posed by deepfakes in the Global South. △ Less

Submitted 19 January, 2026; originally announced January 2026.

How do the Global South Diasporas Mobilize for Transnational Political Change?

Authors: Dipto Das, Afrin Prio, Pritu Saha, Shion Guha, Syed Ishtiaque Ahmed

Abstract: This paper examines how non-resident Bangladeshis mobilized during the 2024 quota-reform turned pro-democracy movement, leveraging social platforms and remittance flows to challenge state authority. Drawing on semi-structured interviews, we identify four phases of their collective action: technology-mediated shifts to active engagement, rapid transnational network building, strategic execution of… ▽ More This paper examines how non-resident Bangladeshis mobilized during the 2024 quota-reform turned pro-democracy movement, leveraging social platforms and remittance flows to challenge state authority. Drawing on semi-structured interviews, we identify four phases of their collective action: technology-mediated shifts to active engagement, rapid transnational network building, strategic execution of remittance boycott, reframing economic dependence as political leverage, and adaptive responses to government surveillance and information blackouts. We extend postcolonial computing by introducing the idea of "diasporic superposition," which shows how diasporas can exercise political and economic influence from hybrid positionalities that both contest and complicate power asymmetries. We reframe diaspora engagement by highlighting how migrants participate in and reshape homeland politics, beyond narratives of integration in host countries. We advance the scholarship on financial technologies by foregrounding their relationship with moral economies of care, state surveillance, regulatory constraints, and uneven international economic power dynamics. Together, these contributions theorize how transnational activism and digital technologies intersect to mobilize political change in Global South contexts. △ Less

Submitted 18 January, 2026; originally announced January 2026.

Digging into the Interior of Hot Cores with ALMA (DIHCA). VI. The Formation of Low-mass Multiple Systems in High-mass Cluster-forming Regions

Authors: Qiuyi Luo, Patricio Sanhueza, Stella S. R. Offner, Fernando Olguin, Adam Ginsburg, Fumitaka Nakamura, Kaho Morii, Yu Cheng, Kei E. I. Tanaka, Junhao Liu, Tie Liu, Xing Lu, Qizhou Zhang, Kotomi Taniguchi, Piyali Saha, Shanghuo Li, Xiaofeng Mai

Abstract: Most stars form in multiple systems, with profound implications in numerous astronomical phenomena intrinsically linked to multiplicity. However, our knowledge about the process on how multiple stellar systems form is incomplete and biased toward nearby molecular clouds forming only low-mass stars, which are unrepresentative of the stellar population in the Galaxy. Most stars form within dense cor… ▽ More Most stars form in multiple systems, with profound implications in numerous astronomical phenomena intrinsically linked to multiplicity. However, our knowledge about the process on how multiple stellar systems form is incomplete and biased toward nearby molecular clouds forming only low-mass stars, which are unrepresentative of the stellar population in the Galaxy. Most stars form within dense cores in clusters alongside high-mass stars (>8 M$_{\odot}$), as likely the Sun did. Here we report deep ALMA 1.33 mm dust continuum observations at ~160 au spatial resolution, revealing 72 low-mass multiple systems embedded in 23 high-mass cluster-forming regions, as part of the Digging into the Interior of Hot Cores with ALMA (DIHCA) survey. We find that the companion separation distribution presents a distinct peak at ~1200 au, in contrast to the one at ~4000 au observed in nearby low-mass regions. The shorter fragmentation scale can be explained by considering the higher pressure exerted by the surrounding medium, which is higher than the one in low-mass regions, due to the larger turbulence and densities involved. Because the peak of the companion separation distribution occurs at much larger scales than the expected disk sizes, we argue that the observed fragmentation is produced by turbulent core fragmentation. Contrary as predicted, the multiplicity fraction remains constant as the stellar density increases. We propose that in the extremely dense environments where high-mass stars form, dynamical interactions play an important role in disrupting weakly bound systems. △ Less

Submitted 17 February, 2026; v1 submitted 13 January, 2026; originally announced January 2026.

Comments: This paper has been accepted by ApJ

MinDist is less than 7

Authors: Purushottam Saha, Diganta Mukherjee

Abstract: The metric MinDist, introduced recently to quantify the distance of an arbitrary Rummy hand from a valid declaration, plays a central role in algorithmic hand evaluation and optimal play. Existing results show that the MinDist of any $13$-card Rummy hand from a single deck is bounded above by $9$. In this paper, we sharpen this bound and prove that the MinDist of any hand is at most $7$. We furthe… ▽ More The metric MinDist, introduced recently to quantify the distance of an arbitrary Rummy hand from a valid declaration, plays a central role in algorithmic hand evaluation and optimal play. Existing results show that the MinDist of any $13$-card Rummy hand from a single deck is bounded above by $9$. In this paper, we sharpen this bound and prove that the MinDist of any hand is at most $7$. We further show that this bound is tight by explicitly exhibiting a hand whose MinDist equals $7$ for a suitable choice of wildcard joker. The proof combines elementary combinatorial arguments with structural properties of card partitions across suits and resolves the gap between the previously known upper bound and the true extremal value. △ Less

Submitted 12 January, 2026; originally announced January 2026.

Comments: 3 pages

MSC Class: 05A99

SPINAL -- Scaling-law and Preference Integration in Neural Alignment Layers

Authors: Arion Das, Partha Pratim Saha, Amit Dhanda, Vinija Jain, Aman Chadha, Amitava Das

Abstract: Direct Preference Optimization (DPO) is a principled, scalable alternative to RLHF for aligning large language models from pairwise preferences, but its internal geometric footprint remains undercharacterized, limiting audits, checkpoint comparisons, and failure prediction. We introduce SPINAL (Scaling-law and Preference Integration in Neural Alignment Layers), a diagnostic that measures how align… ▽ More Direct Preference Optimization (DPO) is a principled, scalable alternative to RLHF for aligning large language models from pairwise preferences, but its internal geometric footprint remains undercharacterized, limiting audits, checkpoint comparisons, and failure prediction. We introduce SPINAL (Scaling-law and Preference Integration in Neural Alignment Layers), a diagnostic that measures how alignment reshapes representations across depth by tracing localized structural change layer by layer. Across model families, DPO produces a layerwise calibration effect concentrated in the final decoder blocks (often layers 21-30), where preference gradients most directly affect the next-token distribution. SPINAL encodes each checkpoint as a depth trace over (layer index, contraction score, transport score). The contraction score summarizes how quickly the tail of a layer's spectrum decays (how fast small modes vanish); higher values indicate stronger contraction into fewer effective directions. The transport score summarizes how much the token distribution shifts between adjacent layers using a bounded overlap measure; lower values indicate shorter, smoother steps through representation space. Aligned checkpoints show a late-layer ramp-up in contraction and a smooth reduction in transport, consistent with tightened and stabilized policy mass, while unaligned models trace higher-curvature, more entropic, and geometrically incoherent depth paths. Overall, alignment is geometrically localized: the final layers encode the dominant preference-induced corrections. SPINAL turns this localization into a practical audit signal, quantifying where alignment concentrates, how strongly it manifests, and when it begins to destabilize during training. △ Less

Submitted 8 January, 2026; originally announced January 2026.

GRIT -- Geometry-Aware PEFT with K-FACPreconditioning, Fisher-Guided Reprojection, andDynamic Rank Adaptation

Authors: Pritish Saha, Chandrav Rajbangshi, Rudra Goyal, Mohit Goyal, Anurag Deo, Biswajit Roy, Ningthoujam Dhanachandra Singh, Raxit Goswami, Amitava Das

Abstract: Parameter-efficient fine-tuning (PEFT) is the default way to adapt LLMs, but widely used LoRA and QLoRA are largely geometry-agnostic: they optimize in fixed, randomly oriented low-rank subspaces with first-order descent, mostly ignoring local loss curvature. This can inflate the effective update budget and amplify drift along weakly constrained directions. We introduce GRIT, a dynamic, curvature-… ▽ More Parameter-efficient fine-tuning (PEFT) is the default way to adapt LLMs, but widely used LoRA and QLoRA are largely geometry-agnostic: they optimize in fixed, randomly oriented low-rank subspaces with first-order descent, mostly ignoring local loss curvature. This can inflate the effective update budget and amplify drift along weakly constrained directions. We introduce GRIT, a dynamic, curvature-aware LoRA procedure that preserves the LoRA parameterization but: (1) preconditions gradients in rank space using K-FAC as a natural-gradient proxy; (2) periodically reprojects the low-rank basis onto dominant Fisher eigendirections to suppress drift; and (3) adapts the effective rank from the spectrum so capacity concentrates where signal resides. Across instruction-following, comprehension, and reasoning benchmarks on LLaMA backbones, GRIT matches or surpasses LoRA and QLoRA while reducing trainable parameters by 46% on average (25--80% across tasks), without practical quality loss across prompt styles and data mixes. To model forgetting, we fit a curvature-modulated power law. Empirically, GRIT yields lower drift and a better updates-vs-retention frontier than strong PEFT-optimizer baselines (Orthogonal-LoRA, IA3, DoRA, Eff-FT, Shampoo). △ Less

Submitted 1 January, 2026; originally announced January 2026.

Quantitative Rule-Based Strategy modeling in Classic Indian Rummy: A Metric Optimization Approach

Authors: Purushottam Saha, Avirup Chakraborty, Sourish Sarkar, Subhamoy Maitra, Diganta Mukherjee, Tridib Mukherjee

Abstract: The 13-card variant of Classic Indian Rummy is a sequential game of incomplete information that requires probabilistic reasoning and combinatorial decision-making. This paper proposes a rule-based framework for strategic play, driven by a new hand-evaluation metric termed MinDist. The metric modifies the MinScore metric by quantifying the edit distance between a hand and the nearest valid configur… ▽ More The 13-card variant of Classic Indian Rummy is a sequential game of incomplete information that requires probabilistic reasoning and combinatorial decision-making. This paper proposes a rule-based framework for strategic play, driven by a new hand-evaluation metric termed MinDist. The metric modifies the MinScore metric by quantifying the edit distance between a hand and the nearest valid configuration, thereby capturing structural proximity to completion. We design a computationally efficient algorithm derived from the MinScore algorithm, leveraging dynamic pruning and pattern caching to exactly calculate this metric during play. Opponent hand-modeling is also incorporated within a two-player zero-sum simulation framework, and the resulting strategies are evaluated using statistical hypothesis testing. Empirical results show significant improvement in win rates for MinDist-based agents over traditional heuristics, providing a formal and interpretable step toward algorithmic Rummy strategy design. △ Less

Submitted 26 December, 2025; originally announced January 2026.

Comments: 9 pages, 6 figures, 2 algorithms

MSC Class: 91(Primary); 05(Secondary)

Investigating Deep Learning Models for Ejection Fraction Estimation from Echocardiography Videos

Authors: Shravan Saranyan, Pramit Saha

Abstract: Left ventricular ejection fraction (LVEF) is a key indicator of cardiac function and plays a central role in the diagnosis and management of cardiovascular disease. Echocardiography, as a readily accessible and non-invasive imaging modality, is widely used in clinical practice to estimate LVEF. However, manual assessment of cardiac function from echocardiograms is time-consuming and subject to con… ▽ More Left ventricular ejection fraction (LVEF) is a key indicator of cardiac function and plays a central role in the diagnosis and management of cardiovascular disease. Echocardiography, as a readily accessible and non-invasive imaging modality, is widely used in clinical practice to estimate LVEF. However, manual assessment of cardiac function from echocardiograms is time-consuming and subject to considerable inter-observer variability. Deep learning approaches offer a promising alternative, with the potential to achieve performance comparable to that of experienced human experts. In this study, we investigate the effectiveness of several deep learning architectures for LVEF estimation from echocardiography videos, including 3D Inception, two-stream, and CNN-RNN models. We systematically evaluate architectural modifications and fusion strategies to identify configurations that maximize prediction accuracy. Models were trained and evaluated on the EchoNet-Dynamic dataset, comprising 10,030 echocardiogram videos. Our results demonstrate that modified 3D Inception architectures achieve the best overall performance, with a root mean squared error (RMSE) of 6.79%. Across architectures, we observe a tendency toward overfitting, with smaller and simpler models generally exhibiting improved generalization. Model performance was also found to be highly sensitive to hyperparameter choices, particularly convolutional kernel sizes and normalization strategies. While this study focuses on echocardiography-based LVEF estimation, the insights gained regarding architectural design and training strategies may be applicable to a broader range of medical and non-medical video analysis tasks. △ Less

Submitted 27 December, 2025; originally announced December 2025.

Analyzing Skill Element in Online Fantasy Cricket

Authors: Sarthak Sarkar, Supratim Das, Purushottam Saha, Diganta Mukherjee, Tridib Mukherjee

Abstract: Online fantasy cricket has emerged as large-scale competitive systems in which participants construct virtual teams and compete based on real-world player performances. This massive growth has been accompanied by important questions about whether outcomes are primarily driven by skill or chance. We develop a statistical framework to assess the role of skill in determining success on these platform… ▽ More Online fantasy cricket has emerged as large-scale competitive systems in which participants construct virtual teams and compete based on real-world player performances. This massive growth has been accompanied by important questions about whether outcomes are primarily driven by skill or chance. We develop a statistical framework to assess the role of skill in determining success on these platforms. We construct and analyze a range of deterministic and stochastic team selection strategies, based on recent form, historical statistics, statistical optimization, and multi-criteria decision making. Strategy performance is evaluated based on points, ranks, and payoff under two contest structures Mega and 4x or Nothing. An extensive comparison between different strategies is made to find an optimal set of strategies. To capture adaptive behavior, we further introduce a dynamic tournament model in which agent populations evolve through a softmax reweighting mechanism proportional to positive payoff realizations. We demonstrate our work by running extensive numerical experiments on the IPL 2024 dataset. The results provide quantitative evidence in favor of the skill element present in online fantasy cricket platforms. △ Less

Submitted 24 December, 2025; originally announced December 2025.

Comments: 12 pages, 2 figures

Towards measuring astrophysical third order correlation functions with the H.E.S.S. optical intensity interferometer

Authors: Andreas Zmija, Gisela Anton, Christopher Ingenhuett, Alison Mitchell, Prasenjit Saha, Pedro Silva Batista, Naomi Vogel, Adrian Zink, Robin Kaiser, Stefan Funk

Abstract: The closure phase, the sum of the three Fourier phases in a telescope triangle, is an important tool in astronomical interferometry, helping to reconstruct the geometries of the observed objects. While already established in amplitude interferometry, for the recently expanding field of intensity interferometers the closure phase enables recovering information of the interferometric phases that are… ▽ More The closure phase, the sum of the three Fourier phases in a telescope triangle, is an important tool in astronomical interferometry, helping to reconstruct the geometries of the observed objects. While already established in amplitude interferometry, for the recently expanding field of intensity interferometers the closure phase enables recovering information of the interferometric phases that are otherwise inaccessible with this technique. To extract the (cosine of) the closure phase ($\cos φ$) in intensity interferometry, third-order correlations between three simultaneously measuring telescopes have to be computed. As the signal-to-noise of such three-photon correlations is too small for current generation intensity interferometers, no third-order correlations of astrophysical targets have been recorded so far. In this paper we present the first measurements of third order correlation functions of two stellar systems, Nunki ($σ$ Sgr) and Dschubba ($δ$ Sco), observed with the H.E.S.S. intensity interferometer in 2023. We show how to isolate the three-photon contribution term from the two-photon contributions, in order to access $\cos φ$. For the observed stellar targets the sensitivity is not high enough to extract closure phase information. To demonstrate that the analysis works well we further extract $\cos φ$ in a laboratory experiment, using the H.E.S.S. intensity interferometer and a pseudo-thermal light source. △ Less

Submitted 15 December, 2025; originally announced December 2025.

Exploiting light coherence in astrophysics

Authors: Vitalii Sliusar, Domenico Della Volpe, Benjamin Garcia, Gilles Koziol, Etienne Lyard, Nicolas Produit, Aramis Raiola, Prasenjit Saha, Lucijana Stanic, Roland Walter

Abstract: The Hanbury Brown-Twiss (HBT) effect, discovered in the 1950s and further developed in the 1960s, was originally used to estimate stellar angular diameters through intensity correlations measured by spatially separated detectors. Further developments started from HBT experiments to exploit quantum bunching of photons in incoherent light sources played foundational role in the development of quantu… ▽ More The Hanbury Brown-Twiss (HBT) effect, discovered in the 1950s and further developed in the 1960s, was originally used to estimate stellar angular diameters through intensity correlations measured by spatially separated detectors. Further developments started from HBT experiments to exploit quantum bunching of photons in incoherent light sources played foundational role in the development of quantum optics. When the two detectors in an HBT experiment are co-located, typically implemented using a beam splitter, a zero-baseline intensity correlation is obtained, which after deconvolution of the detector response function, yields the temporal component of the second-order coherence function. Unlike spatial correlations, this function is independent of the source brightness distribution, or its size, giving direct insight into the properties of the source's emission process itself - photon statistics. Along with photometric and spectral information, the second order coherence function can be used to constrain the emission mechanisms and discriminate between thermal, synchrotron, bremsstrahlung and stimulated emission processes. Evolution of the emission processes would likewise drive changes in the second order coherence. Light coherence information along with multi-wavelength observations, can become a complementary "messenger", carrying internal information about the astronomical source. △ Less

Submitted 13 December, 2025; originally announced December 2025.

Comments: 4 page, 2 figures

Stellar physics at sub-nanoradian angular resolution

Authors: P. Batista, J. Biteau, C. Carlile, J. Cortina, D. Della Volpe, D. Dravins, M. Fiori, S. Funk, W. Guerin, T. Hassan, C. Ingenhütt, I. Jiménez Martínez, R. Kaiser, G. Koziol, O. Lai, Q. Luce, E. Lyard, R. Mirzoyan, A. W. M. Mitchell, A. Nomerotski, N. Produit, A. Raiola, P. Saha, T. Schweizer, V. Sliusar , et al. (14 additional authors not shown)

Abstract: Many stars -- if they could be imaged with enough angular resolution -- would exhibit features expected from theory but not possible to extract from spectra. We may group these by increasing complexity as follows. First, smooth variations in brightness across the surface, resembling solar limb darkening but much more prominent and involving more processes in stars with fast spin or external tides.… ▽ More Many stars -- if they could be imaged with enough angular resolution -- would exhibit features expected from theory but not possible to extract from spectra. We may group these by increasing complexity as follows. First, smooth variations in brightness across the surface, resembling solar limb darkening but much more prominent and involving more processes in stars with fast spin or external tides. Next, there are periodic features: not only oscillations, but also convective cells and starspots, which appear to transit across a star as its spins, and exoplanets that really do transit across the star. Then, there are transients like flares. Current optical interferometers provide synthetic apertures of a few hundred metres and angular resolutions down to about nanoradian ($\simeq 0.2\,$milliarcsecond), enough to resolve some of the above features on the nearest upper main-sequence stars, giants and supergiants. Ongoing projects aims to km-scale synthetic apertures, enough to measure the radius of the nearest white dwarf. In this White Paper we briefly discuss what could be observed with synthetic apertures over $\sim20\,$km -- resolving detail on white dwarfs at the level currently possible on supergiants. △ Less

Submitted 11 December, 2025; originally announced December 2025.

Comments: White paper for the ESO call "What science questions will astronomy need to answer in the 2040s?"

Stellar Multiplicity via Speckle Interferometry with the 3.6 m Devasthal Optical Telescope

Authors: Km Nitu Rai, Neelam Panwar, Jeewan C Pandey, T S Kumar, Subrata Sarangi, Prasenjit Saha

Abstract: Conventional ground-based optical telescopes, even those with large apertures, primarily observe stars, close binaries, and multiple systems as unresolved point sources through photometric measurements. Spectroscopy can identify multiple stellar components within a system, but both techniques are fundamentally limited in resolving stellar surfaces and providing direct angular separations. Although… ▽ More Conventional ground-based optical telescopes, even those with large apertures, primarily observe stars, close binaries, and multiple systems as unresolved point sources through photometric measurements. Spectroscopy can identify multiple stellar components within a system, but both techniques are fundamentally limited in resolving stellar surfaces and providing direct angular separations. Although photometric and spectroscopic observations yield critical information on magnitudes/flux, metallicities, and orbital properties, complementary high-angular-resolution methods are required to constrain additional system characteristics, including angular orbital parameters, model-independent distances, radii, and stellar masses. The limitations of these two methods arise due to the Diffraction Limit of the telescopes and atmospheric turbulence. Speckle Interferometry (SI) is a clever and affordable method for ground-based telescopes to work around atmospheric turbulence. In this work, we utilize the speckle images obtained by the 3.6 m DOT and demonstrate the capability of SI to resolve binary systems, measure their orbital separations, and determine their position angles. For systems with faint companions where conventional analysis fails, we employ Bayesian inference to model speckle patterns and estimate orbital parameters with high precision. These results establish the effective methodology for using a medium-sized, 4-m class telescope like the DOT as a high-resolution stellar interferometer and demonstrate the potential of speckle interferometry as a powerful technique to advance optical interferometric studies within Indian astronomy. △ Less

Submitted 8 December, 2025; originally announced December 2025.

Comments: 13 pages, 8 figure

Emergent Quantum Valley Hall Insulator from Electron Interactions in Transition-Metal Dichalcogenide Heterobilayers

Authors: Palash Saha, Michał Zegrodnik

Abstract: We explore the emergence of topological phases in moiré MoTe$_2$/WSe$_2$ bilayer, highlighting the crucial role of spin-orbit coupling and Coulomb interactions at two holes per moiré unit cell $v = 2$. Our analysis uncovers robust Quantum Valley Hall Insulating (QVHI) phase and reveals that long-range interactions alone can mediate the interlayer electron tunneling, generating topologically nontri… ▽ More We explore the emergence of topological phases in moiré MoTe$_2$/WSe$_2$ bilayer, highlighting the crucial role of spin-orbit coupling and Coulomb interactions at two holes per moiré unit cell $v = 2$. Our analysis uncovers robust Quantum Valley Hall Insulating (QVHI) phase and reveals that long-range interactions alone can mediate the interlayer electron tunneling, generating topologically nontrivial bands even in the absence of the corresponding single-particle hopping. Additionally, we show that in the case of band mixing terms originating both from the interaction and single particle physics a competition between topological states realizing $s$-$wave$ and $p\pm ip$-$wave$ symmetries can appear. Moreover, within the considered theoretical framework, we present that by introducing a small Zeeman field, one can lift the band inversion in one of the valleys. This leads to a Quantum Anomalous Hall Insulating (QAHI) state with the topological gap opening in a single valley and the other being topologically trivial. △ Less

Submitted 2 December, 2025; originally announced December 2025.

Magnetic Fields in Massive Star-forming Regions (MagMaR). VI. Magnetic Field Dragging in the Filamentary High-mass Star-forming Region G35.20--0.74N due to Gravity

Authors: Jihye Hwang, Patricio Sanhueza, Josep Miquel Girart, Ian W. Stephens, Maria T. Beltrán, Chi Yan Law, Qizhou Zhang, Junhao Liu, Paulo Cortés, Fernando A. Olguin, Patrick M. Koch, Fumitaka Nakamura, Piyali Saha, Jia-Wei Wang, Fengwei Xu, Henrik Beuther, Kaho Morii, Manuel Fernández López, Wenyu Jiao, Kee-Tae Kim, Shanghuo Li, Luis A. Zapata, Jongsoo Kim, Spandan Choudhury, Yu Cheng , et al. (5 additional authors not shown)

Abstract: We investigate the magnetic field orientation and strength in the massive star-forming region G35.20-0.74N (G35), using polarized dust emission data obtained with the Atacama Large Millimeter/submillimeter Array (ALMA) as part of the Magnetic fields in Massive star-forming Regions (MagMaR) survey. The G35 region shows a filamentary structure (a length of $\sim$0.1 pc) with six bright cores located… ▽ More We investigate the magnetic field orientation and strength in the massive star-forming region G35.20-0.74N (G35), using polarized dust emission data obtained with the Atacama Large Millimeter/submillimeter Array (ALMA) as part of the Magnetic fields in Massive star-forming Regions (MagMaR) survey. The G35 region shows a filamentary structure (a length of $\sim$0.1 pc) with six bright cores located along the filament's long axis. Magnetic field strengths across the G35 region range from 0.2 to 4.4 mG with a mean value of 0.8 $\pm$ 0.4 mG. The mass-to-flux ratio ($λ$) varies from 0.1 to 6.0 the critical value. The highest values are found locally around cores, whereas the remains of the filament are subcritical. A H$^{13}$CO$^+$ (3--2) velocity gradient of 29 km s$^{-1}$ pc$^{-1}$ is evident along the filament's long axis, aligned with the magnetic field direction. At larger scales ($\sim$0.1 pc), the magnetic field lines appear roughly perpendicular to the filament's long axis, in contrast to the smaller-scale structure ($\sim$0.003 pc) traced by ALMA. The magnetic field lines could be dragged along the filament as a result of the gas motion induced by the gravitational potential of the filament. Six cores in the filament have similar spacings between 0.02--0.04 pc. The initial filament fragmentation could have produced a core spacing of 0.06 pc, following filament fragmentation theory, and the current core spacing is the result of cores comoving with the gas along the filament. This core migration could occur in a few 10$^4$ years, consistent with high-mass star formation time scales. △ Less

Submitted 28 October, 2025; originally announced October 2025.

Comments: submitted to AJ

FedAgentBench: Towards Automating Real-world Federated Medical Image Analysis with Server-Client LLM Agents

Authors: Pramit Saha, Joshua Strong, Divyanshu Mishra, Cheng Ouyang, J. Alison Noble

Abstract: Federated learning (FL) allows collaborative model training across healthcare sites without sharing sensitive patient data. However, real-world FL deployment is often hindered by complex operational challenges that demand substantial human efforts. This includes: (a) selecting appropriate clients (hospitals), (b) coordinating between the central server and clients, (c) client-level data pre-proces… ▽ More Federated learning (FL) allows collaborative model training across healthcare sites without sharing sensitive patient data. However, real-world FL deployment is often hindered by complex operational challenges that demand substantial human efforts. This includes: (a) selecting appropriate clients (hospitals), (b) coordinating between the central server and clients, (c) client-level data pre-processing, (d) harmonizing non-standardized data and labels across clients, and (e) selecting FL algorithms based on user instructions and cross-client data characteristics. However, the existing FL works overlook these practical orchestration challenges. These operational bottlenecks motivate the need for autonomous, agent-driven FL systems, where intelligent agents at each hospital client and the central server agent collaboratively manage FL setup and model training with minimal human intervention. To this end, we first introduce an agent-driven FL framework that captures key phases of real-world FL workflows from client selection to training completion and a benchmark dubbed FedAgentBench that evaluates the ability of LLM agents to autonomously coordinate healthcare FL. Our framework incorporates 40 FL algorithms, each tailored to address diverse task-specific requirements and cross-client characteristics. Furthermore, we introduce a diverse set of complex tasks across 201 carefully curated datasets, simulating 6 modality-specific real-world healthcare environments, viz., Dermatoscopy, Ultrasound, Fundus, Histopathology, MRI, and X-Ray. We assess the agentic performance of 14 open-source and 10 proprietary LLMs spanning small, medium, and large model scales. While some agent cores such as GPT-4.1 and DeepSeek V3 can automate various stages of the FL pipeline, our results reveal that more complex, interdependent tasks based on implicit goals remain challenging for even the strongest models. △ Less

Submitted 28 September, 2025; originally announced September 2025.

First results from ALPPS: a sub-Alfvénic streamer in SVS13A

Authors: P. C. Cortes, J. E. Pineda, T. -H. Hsieh, J. J. Tobin, P. Saha, J. M. Girart, V. J. M. Le Gouellec, I. W. Stephens, L. W. Looney, E. Koumpia, M. T. Valdivia-Mena, L. Cacciapuoti, C. Gieser, S. S. R. Offner, P. Caselli, P. Sanhueza, D. Segura-Cox, M. Fernandez-Lopez, K. Morii, B. Huang, F. O. Alves, Q. Zhang, W. Kwon, C. L. H. Hull, Z. Y. Li

Abstract: We present the first results from the ALMA Perseus Polarization Survey (ALPPS), focusing on the magnetic field in the SVS13A circumbinary disk. The dataset includes full-Stokes dust continuum observations at $\sim0\farcs3$ and 870 $μ$m, as well as molecular line emission from C$^{17}$O$(J=3 \rightarrow 2)$ at $\sim0\farcs3$, C$^{18}$O$(J=2 \rightarrow 1)$ at $\sim0\farcs2$, and DCN… ▽ More We present the first results from the ALMA Perseus Polarization Survey (ALPPS), focusing on the magnetic field in the SVS13A circumbinary disk. The dataset includes full-Stokes dust continuum observations at $\sim0\farcs3$ and 870 $μ$m, as well as molecular line emission from C$^{17}$O$(J=3 \rightarrow 2)$ at $\sim0\farcs3$, C$^{18}$O$(J=2 \rightarrow 1)$ at $\sim0\farcs2$, and DCN$(J=3 \rightarrow 2)$ at $\sim0\farcs1$ angular resolution. Our observations resolve both a previously identified dust spiral and an infalling streamer, capturing their spatial and kinematic structures. The streamer is traced from scales $>300$ au down to the circumbinary disk. Using alignment measure (AM) maps and histograms that compare the orientations of the plane-of-sky magnetic field with local intensity and velocity gradients, we find that the AM distribution peaks at a value of 1. This AM peak strongly suggests alignment between the field and the dust total intensity emission, as well as between the field and the gas velocity, which in turn suggests grain alignment by magnetic fields. From our data, we derive a magnetic field strength, B$_{\mathrm{pos}} \sim 1.1 \pm 0.6$\, mG, and a kinetic to magnetic energy ratio of $0.5 \pm 0.4$, suggesting magnetic dominance. We also produced a map of the Alfvénic Mach number, finding $\mathcal{M}_{\rm A} < 1$ along the streamer, consistent with sub-Alfvénic infalling motions. Therefore, the field is likely facilitating the inflow of material from the envelope onto the disk by constraining movement across the field lines. This represents the first detection of a magnetically sub-Alfvénic infalling streamer in a protostellar system. △ Less

Submitted 25 September, 2025; originally announced September 2025.

Comments: Accepted for publication is Astrophiscal Journal Letters

Probing Stellar Kinematics with the Time-Asymmetric Hanbury Brown and Twiss Effect

Authors: Lucijana Stanic, Ivan Cardea, Edoardo Charbon, Domenico Della Volpe, Daniel Florin, Andrea Guerrieri, Gilles Koziol, Etienne Lyard, Nicolas Produit, Aramis Raiola, Prasenjit Saha, Vitalii Sliusar, Achim Vollhardt, Roland Walter

Abstract: Intensity interferometry (II) offers a powerful means to observe stellar objects with a high resolution. In this work, we demonstrate that II can also probe internal stellar kinematics by revealing a time-asymmetric Hanbury Brown and Twiss (HBT) effect, causing a measurable shift in the temporal correlation peak away from zero delay. We develop numerical models to simulate this effect for two dist… ▽ More Intensity interferometry (II) offers a powerful means to observe stellar objects with a high resolution. In this work, we demonstrate that II can also probe internal stellar kinematics by revealing a time-asymmetric Hanbury Brown and Twiss (HBT) effect, causing a measurable shift in the temporal correlation peak away from zero delay. We develop numerical models to simulate this effect for two distinct astrophysical scenarios: an emission-line circumstellar disk and an absorption-line binary system. Our simulations reveal a clear sensitivity of this temporal asymmetry to the system's inclination angle, velocity symmetry, and internal dynamics. This suggests that, with sufficiently high time resolution, II can be used to extract quantitative information about internal kinematics, offering a new observational window on stellar dynamics. △ Less

Submitted 26 January, 2026; v1 submitted 16 September, 2025; originally announced September 2025.

Comments: 11 pages, 6 figures, accepted in OJAp

Journal ref: The Open Journal of Astrophysics Vol. 9 (May) 2025

Neural Collapse-Inspired Multi-Label Federated Learning under Label-Distribution Skew

Authors: Can Peng, Yuyuan Liu, Yingyu Yang, Pramit Saha, Qianye Yang, J. Alison Noble

Abstract: Federated Learning (FL) enables collaborative model training across distributed clients while preserving data privacy, but remains challenging when client data are highly heterogeneous. These challenges are further amplified in multi-label scenarios, where inter-label dependencies and mismatches between local and global label relationships introduce additional optimization conflicts. While most FL… ▽ More Federated Learning (FL) enables collaborative model training across distributed clients while preserving data privacy, but remains challenging when client data are highly heterogeneous. These challenges are further amplified in multi-label scenarios, where inter-label dependencies and mismatches between local and global label relationships introduce additional optimization conflicts. While most FL studies focus on single-label classification, many real-world applications are inherently multi-label and often exhibit severe label skew across clients. To address this important yet underexplored problem, we propose FedNCA-ML, a novel FL framework that aligns client representations and learns discriminative, well-clustered features inspired by Neural Collapse (NC) theory. NC describes an ideal latent geometry where each class's features collapse to their mean, forming a maximally separated simplex. FedNCA-ML further introduces an attention-based module to extract class-specific representations, enabling more balanced learning under heavy label imbalance. These class-wise representations are then aligned via a shared NC-inspired structure, mitigating inter-client conflicts induced by heterogeneous local data and inconsistent label dependencies. In addition, we design regularisation losses to encourage compact and consistent feature clustering in the latent space. Experiments on five benchmark datasets under nine FL settings demonstrate the effectiveness of the proposed method, achieving improvements of up to 3.92% in class-wise AUC and 4.93% in class-wise F1 score. △ Less

Submitted 22 March, 2026; v1 submitted 15 September, 2025; originally announced September 2025.

A Holistic Approach to E-Commerce Innovation: Redefining Security and User Experience

Authors: Mohammad Olid Ali Akash, Priyangana Saha

Abstract: In the modern, fast-moving world of e-commerce, many Android apps face challenges in providing a simple and secure shopping experience. Many of these apps, often enough, have complicated designs that prevent users from finding what they want quickly, thus frustrating them and wasting their precious time. Another major issue is that of security; with the limitation of payment options and weak authe… ▽ More In the modern, fast-moving world of e-commerce, many Android apps face challenges in providing a simple and secure shopping experience. Many of these apps, often enough, have complicated designs that prevent users from finding what they want quickly, thus frustrating them and wasting their precious time. Another major issue is that of security; with the limitation of payment options and weak authentication mechanisms, users' sensitive information can be compromised. This research presents a new e-commerce platform that responds to the above challenges with an intuitive interface and strong security measures. The platform makes online shopping easy with well-organized categories of products and a fast, efficient checkout process. It also gives priority to security by incorporating features such as Google authentication and SSL-secured payment gateways to protect user data and ensure secure transactions. This paper discusses how a focus on user-friendliness, security, and personalization steps up the game for e-commerce platforms, providing workable frameworks that match modern user needs and expectations. The findings show the e-commerce user experience can be remodelled by the platform, hence opening ways for future developments in that respect. △ Less

Submitted 15 September, 2025; originally announced September 2025.

Prompting Away Stereotypes? Evaluating Bias in Text-to-Image Models for Occupations

Authors: Shaina Raza, Maximus Powers, Partha Pratim Saha, Mahveen Raza, Rizwan Qureshi

Abstract: Text-to-Image (TTI) models are powerful creative tools but risk amplifying harmful social biases. We frame representational societal bias assessment as an image curation and evaluation task and introduce a pilot benchmark of occupational portrayals spanning five socially salient roles (CEO, Nurse, Software Engineer, Teacher, Athlete). Using five state-of-the-art models: closed-source (DALLE 3, Gem… ▽ More Text-to-Image (TTI) models are powerful creative tools but risk amplifying harmful social biases. We frame representational societal bias assessment as an image curation and evaluation task and introduce a pilot benchmark of occupational portrayals spanning five socially salient roles (CEO, Nurse, Software Engineer, Teacher, Athlete). Using five state-of-the-art models: closed-source (DALLE 3, Gemini Imagen 4.0) and open-source (FLUX.1-dev, Stable Diffusion XL Turbo, Grok-2 Image), we compare neutral baseline prompts against fairness-aware controlled prompts designed to encourage demographic diversity. All outputs are annotated for gender (male, female) and race (Asian, Black, White), enabling structured distributional analysis. Results show that prompting can substantially shift demographic representations, but with highly model-specific effects: some systems diversify effectively, others overcorrect into unrealistic uniformity, and some show little responsiveness. These findings highlight both the promise and the limitations of prompting as a fairness intervention, underscoring the need for complementary model-level strategies. We release all code and data for transparency and reproducibility https://github.com/maximus-powers/img-gen-bias-analysis. △ Less

Submitted 31 August, 2025; originally announced September 2025.

Furstenberg--Sárközy theorem over number fields

Authors: Dev Ranjan Pandey, Jyoti Prakash Saha

Abstract: We introduce the notion of intersective polynomials having coefficients in the ring of integers $\mathscr{O}_K$ of a number field $K$, and define a notion of upper density of subsets of $\mathscr{O}_K$. We prove that given any intersective polynomial $p(x)$ over $\mathscr{O}_K$, every subset $A$ of $\mathscr{O}_K$ of positive upper density contains two distinct elements whose difference is equal t… ▽ More We introduce the notion of intersective polynomials having coefficients in the ring of integers $\mathscr{O}_K$ of a number field $K$, and define a notion of upper density of subsets of $\mathscr{O}_K$. We prove that given any intersective polynomial $p(x)$ over $\mathscr{O}_K$, every subset $A$ of $\mathscr{O}_K$ of positive upper density contains two distinct elements whose difference is equal to $p(x)$ for some element $x$ in $\mathscr{O}_K$. Moreover, we obtain a quantitative version of this result. The proof is motivated by an argument due to Lucier, and the Fourier-free proof of the Furstenberg--Sárközy theorem over the integers by Green, Tao and Ziegler. △ Less

Submitted 8 October, 2025; v1 submitted 26 August, 2025; originally announced August 2025.

MSC Class: 11B30; 05B10

A Study of Privacy-preserving Language Modeling Approaches

Authors: Pritilata Saha, Abhirup Sinha

Abstract: Recent developments in language modeling have increased their use in various applications and domains. Language models, often trained on sensitive data, can memorize and disclose this information during privacy attacks, raising concerns about protecting individuals' privacy rights. Preserving privacy in language models has become a crucial area of research, as privacy is one of the fundamental hum… ▽ More Recent developments in language modeling have increased their use in various applications and domains. Language models, often trained on sensitive data, can memorize and disclose this information during privacy attacks, raising concerns about protecting individuals' privacy rights. Preserving privacy in language models has become a crucial area of research, as privacy is one of the fundamental human rights. Despite its significance, understanding of how much privacy risk these language models possess and how it can be mitigated is still limited. This research addresses this by providing a comprehensive study of the privacy-preserving language modeling approaches. This study gives an in-depth overview of these approaches, highlights their strengths, and investigates their limitations. The outcomes of this study contribute to the ongoing research on privacy-preserving language modeling, providing valuable insights and outlining future research directions. △ Less

Submitted 21 August, 2025; originally announced August 2025.

Predicting Road Crossing Behaviour using Pose Detection and Sequence Modelling

Authors: Subhasis Dasgupta, Preetam Saha, Agniva Roy, Jaydip Sen

Abstract: The world is constantly moving towards AI based systems and autonomous vehicles are now reality in different parts of the world. These vehicles require sensors and cameras to detect objects and maneuver according to that. It becomes important to for such vehicles to also predict from a distant if a person is about to cross a road or not. The current study focused on predicting the intent of crossi… ▽ More The world is constantly moving towards AI based systems and autonomous vehicles are now reality in different parts of the world. These vehicles require sensors and cameras to detect objects and maneuver according to that. It becomes important to for such vehicles to also predict from a distant if a person is about to cross a road or not. The current study focused on predicting the intent of crossing the road by pedestrians in an experimental setup. The study involved working with deep learning models to predict poses and sequence modelling for temporal predictions. The study analysed three different sequence modelling to understand the prediction behaviour and it was found out that GRU was better in predicting the intent compared to LSTM model but 1D CNN was the best model in terms of speed. The study involved video analysis, and the output of pose detection model was integrated later on to sequence modelling techniques for an end-to-end deep learning framework for predicting road crossing intents. △ Less

Submitted 21 August, 2025; originally announced August 2025.

Comments: This is a pre-print version of the original paper accepted in the IEEE conference INDISCON 2025. It contains 8 figures and 1 table. The length of the paper is 7 pages

Radial Pulsations in Polaris: A Secondary Science Application of Cherenkov Telescopes via Intensity Interferometry

Authors: Km Nitu Rai, Prasenjit Saha, Subrata Sarangi

Abstract: Ground-based Cherenkov telescopes, although typically inoperative during moonlit nights for gamma-ray observations, offer a valuable opportunity for secondary scientific applications through Intensity Interferometry (II). Recent developments and observations suggest that implementing II instrumentation on existing Imaging Atmospheric Cherenkov Telescopes (IACTs) or the Cherenkov Telescope Array (C… ▽ More Ground-based Cherenkov telescopes, although typically inoperative during moonlit nights for gamma-ray observations, offer a valuable opportunity for secondary scientific applications through Intensity Interferometry (II). Recent developments and observations suggest that implementing II instrumentation on existing Imaging Atmospheric Cherenkov Telescopes (IACTs) or the Cherenkov Telescope Array (CTA) can significantly advance optical stellar measurements. Motivated by the resurgence of II efforts over the past two decades, this work presents simulations demonstrating the estimation of stellar parameters for a radially pulsating star, such as Polaris, using either a single telescope or multiple telescopes. For single-telescope simulations, we assume that the photon pixels in the camera are mapped onto four distinct regions of the aperture, generating multiple baselines and enabling enhanced observational plane coverage. These results highlight the potential of Cherenkov telescopes in India for high-resolution optical astronomy during otherwise inoperative periods and offer promising insights into the characterization of bright stellar objects with unprecedented precision. △ Less

Submitted 13 August, 2025; originally announced August 2025.

Comments: 10 pages, 7 figures

Generative AI for image reconstruction in Intensity Interferometry: a first attempt

Authors: Km Nitu Rai, Yuri van der Burg, Soumen Basak, Prasenjit Saha, Subrata Sarangi

Abstract: In the last few years Intensity Interferometry (II) has made significant strides in achieving high-precision resolution of stellar objects at optical wavelengths. Despite these advancements, phase retrieval remains a major challenge due to the nature of photon correlation. This paper explores the application of a conditional Generative Adversarial Network (cGAN) to tackle the problem of image reco… ▽ More In the last few years Intensity Interferometry (II) has made significant strides in achieving high-precision resolution of stellar objects at optical wavelengths. Despite these advancements, phase retrieval remains a major challenge due to the nature of photon correlation. This paper explores the application of a conditional Generative Adversarial Network (cGAN) to tackle the problem of image reconstruction in Intensity Interferometry. This approach successfully reconstructs the shape, size, and brightness distribution of a fast-rotating star from sparsely sampled, spatial power spectrum of the source, corresponding to II with four telescopes. Although this particular example could also be addressed using parameter fitting, the results suggest that with larger arrays much more complicated systems could be reconstructed by applying machine-learning techniques to II. △ Less

Submitted 8 August, 2025; v1 submitted 5 August, 2025; originally announced August 2025.

Comments: 14 pages, 16 figures

Current State in Privacy-Preserving Text Preprocessing for Domain-Agnostic NLP

Authors: Abhirup Sinha, Pritilata Saha, Tithi Saha

Abstract: Privacy is a fundamental human right. Data privacy is protected by different regulations, such as GDPR. However, modern large language models require a huge amount of data to learn linguistic variations, and the data often contains private information. Research has shown that it is possible to extract private information from such language models. Thus, anonymizing such private and sensitive infor… ▽ More Privacy is a fundamental human right. Data privacy is protected by different regulations, such as GDPR. However, modern large language models require a huge amount of data to learn linguistic variations, and the data often contains private information. Research has shown that it is possible to extract private information from such language models. Thus, anonymizing such private and sensitive information is of utmost importance. While complete anonymization may not be possible, a number of different pre-processing approaches exist for masking or pseudonymizing private information in textual data. This report focuses on a few of such approaches for domain-agnostic NLP tasks. △ Less

Submitted 5 August, 2025; originally announced August 2025.

Comments: To be published in the Proceedings of Die Studierendenkonferenz Informatik (SKILL) 2024

Gravitational lensing of fast radio bursts: prospects for probing microlens populations in lensing galaxies

Authors: Ashish Kumar Meena, Prasenjit Saha

Abstract: Gravitational lensing by a stellar microlens of mass $M$ forms two images separated by micro-arcseconds on the sky and has a time delay of $2\times10^{-5}(M/{\rm M_\odot})$ seconds. Although we cannot resolve such micro-images in the sky, they could be resolved in time if the source is a fast radio burst (FRB). In this work, we study the magnification ($|μ|$) and time delay~($t_d$) distributions o… ▽ More Gravitational lensing by a stellar microlens of mass $M$ forms two images separated by micro-arcseconds on the sky and has a time delay of $2\times10^{-5}(M/{\rm M_\odot})$ seconds. Although we cannot resolve such micro-images in the sky, they could be resolved in time if the source is a fast radio burst (FRB). In this work, we study the magnification ($|μ|$) and time delay~($t_d$) distributions of micro-images led by different microlens populations. We find that, in microlensing of typical strongly lensed (macro-)images in galaxy lenses, micro-images stemmed from a population of stellar mass microlenses in the $[0.08, 1.5]\:{\rm M_\odot}$ range and a second (dark) microlens population in $[10^{-3} - 10^{-2}]\:{\rm M_\odot}$ range reside in different parts of $|μ|-t_d$ plane. For the global minimum macro-image, due to low stellar mass density, we find that the stellar population leads to peaks in autocorrelation at ${>}10^{-6}$ seconds, whereas the secondary population leads to peaks at ${<}10^{-6}$ seconds, allowing us to differentiate different microlens populations. However, an increase in stellar density introduces new peaks at ${<}10^{-6}$ seconds, which can pollute the inference about the presence of multiple microlens populations. In addition, we also show that the number of micro-images, hence the number of peaks in the autocorrelation, is also sensitive to the underlying stellar mass function, allowing us to constrain the stellar initial mass function (IMF) with FRB microlesning in the future. This work is a first step towards using FRB lensing to probe the microlens population within strong lenses, and more detailed studies are required to assess the effect of various uncertainties that we only discussed qualitatively. △ Less

Submitted 20 November, 2025; v1 submitted 27 July, 2025; originally announced July 2025.

Comments: 16 pages. 6 figures. 1 table. Accepted in Phys Rev D

Journal ref: Phys Rev D, 112, 123012 (2025)

Reconstructing Biological Pathways by Applying Selective Incremental Learning to (Very) Small Language Models

Authors: Pranta Saha, Joyce Reimer, Brook Byrns, Connor Burbridge, Neeraj Dhar, Jeffrey Chen, Steven Rayan, Gordon Broderick

Abstract: The use of generative artificial intelligence (AI) models is becoming ubiquitous in many fields. Though progress continues to be made, general purpose large language AI models (LLM) show a tendency to deliver creative answers, often called "hallucinations", which have slowed their application in the medical and biomedical fields where accuracy is paramount. We propose that the design and use of mu… ▽ More The use of generative artificial intelligence (AI) models is becoming ubiquitous in many fields. Though progress continues to be made, general purpose large language AI models (LLM) show a tendency to deliver creative answers, often called "hallucinations", which have slowed their application in the medical and biomedical fields where accuracy is paramount. We propose that the design and use of much smaller, domain and even task-specific LM may be a more rational and appropriate use of this technology in biomedical research. In this work we apply a very small LM by today's standards to the specialized task of predicting regulatory interactions between molecular components to fill gaps in our current understanding of intracellular pathways. Toward this we attempt to correctly posit known pathway-informed interactions recovered from manually curated pathway databases by selecting and using only the most informative examples as part of an active learning scheme. With this example we show that a small (~110 million parameters) LM based on a Bidirectional Encoder Representations from Transformers (BERT) architecture can propose molecular interactions relevant to tuberculosis persistence and transmission with over 80% accuracy using less than 25% of the ~520 regulatory relationships in question. Using information entropy as a metric for the iterative selection of new tuning examples, we also find that increased accuracy is driven by favoring the use of the incorrectly assigned statements with the highest certainty (lowest entropy). In contrast, the concurrent use of correct but least certain examples contributed little and may have even been detrimental to the learning rate. △ Less

Submitted 6 July, 2025; originally announced July 2025.

Comments: 9 pages, 6 figures, 3 tables + 28 pages of supplemental tables; submitted to 16th ACM Conference on Bioinformatics, Computational Biology, and Health Informatics (ACM BCB 2025) as submission no. 76

HatePRISM: Policies, Platforms, and Research Integration. Advancing NLP for Hate Speech Proactive Mitigation

Authors: Naquee Rizwan, Seid Muhie Yimam, Daryna Dementieva, Florian Skupin, Tim Fischer, Daniil Moskovskiy, Aarushi Ajay Borkar, Robert Geislinger, Punyajoy Saha, Sarthak Roy, Martin Semmann, Alexander Panchenko, Chris Biemann, Animesh Mukherjee

Abstract: Despite regulations imposed by nations and social media platforms, e.g. (Government of India, 2021; European Parliament and Council of the European Union, 2022), inter alia, hateful content persists as a significant challenge. Existing approaches primarily rely on reactive measures such as blocking or suspending offensive messages, with emerging strategies focusing on proactive measurements like d… ▽ More Despite regulations imposed by nations and social media platforms, e.g. (Government of India, 2021; European Parliament and Council of the European Union, 2022), inter alia, hateful content persists as a significant challenge. Existing approaches primarily rely on reactive measures such as blocking or suspending offensive messages, with emerging strategies focusing on proactive measurements like detoxification and counterspeech. In our work, which we call HatePRISM, we conduct a comprehensive examination of hate speech regulations and strategies from three perspectives: country regulations, social platform policies, and NLP research datasets. Our findings reveal significant inconsistencies in hate speech definitions and moderation practices across jurisdictions and platforms, alongside a lack of alignment with research efforts. Based on these insights, we suggest ideas and research direction for further exploration of a unified framework for automated hate speech moderation incorporating diverse strategies. △ Less

Submitted 6 July, 2025; originally announced July 2025.

Comments: arXiv admin note: substantial text overlap with arXiv:2406.19543

GPU-accelerated Modeling of Biological Regulatory Networks

Authors: Joyce Reimer, Pranta Saha, Chris Chen, Neeraj Dhar, Brook Byrns, Steven Rayan, Gordon Broderick

Abstract: The complex regulatory dynamics of a biological network can be succinctly captured using discrete logic models. Given even sparse time-course data from the system of interest, previous work has shown that global optimization schemes are suitable for proposing logic models that explain the data and make predictions about how the system will behave under varying conditions. Considering the large sca… ▽ More The complex regulatory dynamics of a biological network can be succinctly captured using discrete logic models. Given even sparse time-course data from the system of interest, previous work has shown that global optimization schemes are suitable for proposing logic models that explain the data and make predictions about how the system will behave under varying conditions. Considering the large scale of the parameter search spaces associated with these regulatory systems, performance optimizations on the level of both hardware and software are necessary for making this a practical tool for in silico pharmaceutical research. We show here how the implementation of these global optimization algorithms in a GPU-computing environment can accelerate the solution of these parameter search problems considerably. We carry out parameter searches on two model biological regulatory systems that represent almost an order of magnitude scale-up in complexity, and we find the gains in efficiency from GPU to be a 33%-43% improvement compared to multi-thread CPU implementations and a 33%-1866% increase compared to CPU in serial. These improvements make global optimization of logic model identification a far more attractive and feasible method for in silico hypothesis generation and design of experiments. △ Less

Submitted 10 June, 2025; originally announced June 2025.

Comments: 10 pages, 5 figures, 2 tables; submitted to 16th ACM Conference on Bioinformatics, Computational Biology, and Health Informatics (ACM BCB 2025) as submission no. 6