arXiv subjects

Supriyo Ghosh

Publications and source records attributed to Supriyo Ghosh.

At least 19 recordsLinked to original sources

arXiv2026

Flat band and Bulk-Boundary correspondence in a non-Hermitian trimerized lattice model with generic boundary conditions

We consider a Su-Schrieffer-Heeger(SSH)-type trimer model with next-nearest-neighbor(NNN) interaction and balanced loss-gain(BLG) to study the combined effect of lattice symmetries, topology, non-hermiticity and general boundary conditions(GBC)on the existence of flat band and the nature of Bulk-Boundary correspondence(BBC). We derive the necessary and sufficient conditions for the existence of an entirely real spectrum under the periodic boundary condition(PBC). The exact expressions for the compact localized states(CLS) and energy eigenvalues corresponding to flat bands are derived analytically under the PBC. We establish topological phase transitions(TPT) for PT-symmetry and pseudo-chiral symmetry through the computation of the Zak phase and sub-lattice Zak phase, respectively. The Hamiltonian under the open boundary condition(OBC) is studied numerically, and edge states are observed in the topologically non-trivial phase, thereby establishing the non-hermitian BBC. The CLS exists in both bulk and the boundary for systems having only pseudo-chiral symmetry, and an additional PT-symmetry destroys the CLS at the boundary. We generalize a known formalism to study the same Hamiltonian under GBC, and derive analytic expressions for the energy and eigenstates for a class of boundary conditions in parametric ranges which admit flat band under the PBC. The edge states for these boundary conditions, including the OBC, are obtained analytically in the topologically non-trivial phase, thereby establishing BBC. The non-hermitian skin effect(NHSE) is seen in the model with reciprocal bulk interaction and strongly non-reciprocal boundary terms. The winding number based on spectral topology is computed analytically.

cond-mat.mes-hall↗

arXiv2026

PETRA: Transforming Web Text for Petroleum-Engineering Domain Adaptation

Petroleum-engineering search exposes a supervision gap for strong general retrievers: relevant evidence exists in public web text, but domain relevance labels are scarce. To address this gap, we propose PETRA, a large-scale Petroleum Engineering Text for Retrieval Adaptation dataset and pipeline that converts noisy public web data into a curated domain corpus and synthetic supervision for dense retrieval and reranking. PETRA contains 1.36M curated chunks, approximately 2B token equivalents, $\approx$859k, embedding training rows from $\approx$224k anchors, and roughly 400k teacher-scored reranker candidate rows. Its construction combines high-recall energy-domain curation, an energy-domain classifier with 98.4% test accuracy, chunk-grounded query generation, LLM-written hard negatives, and retrieval-mined candidate lists. PETRA improves first-stage in-domain Normalized Discounted Cumulative Gain (nDCG) from 0.703 to 0.763 through score fusion. Reranker adaptation improves the public Earth Science benchmark by 44% relative and a six-task reasoning-intensive panel by 23%. Failed training recipes show that high train-holdout accuracy on synthetic labels does not predict retrieval gains; retrieval-mined data helps only after being repackaged as teacher-scored candidate lists sampled from the inference-time candidate distribution.

cs.IR↗

arXiv2026

Magnetic HIP-NN for spin dynamics in disordered itinerant magnets

We present a magnetic extension of the Hierarchically Interacting Particle Neural Network (HIP-NN) that enables large-scale simulations of electron-mediated spin dynamics in disordered itinerant magnets. The resulting magnetic HIP-NN (mHIP-NN) incorporates rotationally invariant spin correlations directly into hierarchical message-passing layers, enabling the network to learn emergent magnetic energy landscapes and effective local fields from coupled geometric-spin environments while preserving spin-rotation symmetry. As a benchmark application, we consider structurally disordered itinerant $s$-$d$ exchange models in which the effective magnetic forces arise dynamically from the instantaneous electronic structure and are computationally prohibitive to evaluate using conventional exact-diagonalization-based approaches. We show that mHIP-NN accurately reproduces the local torques governing Landau-Lifshitz-Gilbert dynamics and faithfully captures the nonequilibrium evolution of spatial spin correlations following thermal quenches. Our results establish symmetry-aware hierarchical message-passing networks as an efficient and scalable framework for large-scale simulations of frustrated itinerant spin systems and nonequilibrium magnetic dynamics. More broadly, because the learned energy functional remains fully differentiable with respect to both atomic coordinates and spin variables, the framework also provides a natural foundation for spin-dependent interatomic potentials and coupled atom-spin dynamics.

cond-mat.dis-nn↗

arXiv2026

Dynamic Range Beyond Bit Depth of a CMOS Image Sensor Using Interleaved Row Readout

The dynamic range available from a sensor is vital to its utility. The limits on the dynamic range that can be obtained from an image sensor are set by the brightest and faintest objects that can be detected. In recent years, CMOS (complementary metal-oxide-semiconductor) image sensors (CIS) have gained high popularity due to their low cost and high availability. However, as with all detectors, the dynamic range is constrained by the sensor's bit depth. Here, we have modified the readout scheme of a commercial CIS120 sensor from Teledyne e2v, to enhance its dynamic range. We have advanced the interleaved row readout method proposed by Wocial et al. by using a more sophisticated approach, which enables us to readout chosen rows much more frequently to avoid saturation and then readout other rows on the sensor once to form the image. Our laboratory tests provide a dynamic range of 134 dB elevated from the sensor's native 12-bit of about 71 dB. We also built a camera housing that enabled first-time operation of interleaved row readout on-sky to observe the bright stars, Vega and Polaris. In complex mode we obtained unsaturated single exposure images of these bright stars, which have magnitudes near zero and detect background stars with Gaia G magnitudes around 15 in a single exposure, with a detection threshold of 5$\sigma$. The achievable dynamic range with this interleaved row readout is limited only by the readout noise and scattering in the camera optics.

astro-ph.IM↗

arXiv2026

FixItFlow: Automated Troubleshooting Guide Generation from Cloud Incidents

Cloud services experience frequent incidents that require rapid diagnosis and resolution. Troubleshooting guides help engineers respond consistently, but creating them manually is labor-intensive, resulting in incomplete coverage and outdated documentation. We present FixItFlow, an automated system that generates troubleshooting guides from historical incident data using large language models. The system extracts diagnostic patterns from engineer actions, synthesizes structured guides with verified commands, and enforces strict validation to prevent fabricated content. In our evaluation with 26 engineers, generated guides achieved 61.5\% positive ratings for clarity and demonstrated a 2.3x reduction in mitigation time for incidents with associated guides. These results indicate that automated guide generation can improve incident response while reducing documentation burden on engineering teams.

cs.CL↗

arXiv2026

pyTANSPEC v1.0 and HxRGproc: Updated packages to Clean and Reduce TANSPEC data

TIFR-ARIES Near-Infrared Spectrometer (TANSPEC) is a spectrograph-cum-imager operating over the wavelength range $0.55 - 2.5~\mu$m. The instrument is mounted on the 3.6-m Devasthal Optical Telescope (3.6-m DOT). It offers two resolution modes: Low Resolution (LR) with $R\sim100-350$ and Cross-Dispersed (XD) via various slits of different widths (0.5", 0.75", 1.0", 1.5", 2.0" and 4.0"). The LR mode provides a resolving power ($R$) of $\sim 100-350$, while the XD mode achieves $R\sim2500$ using the 0.5" slit. The previous version of the data reduction pipeline supported only wavelength-calibrated XD mode spectra and was limited to two slits (S-0.5 and S-1.0). In this work, we present an upgraded version of pyTANSPEC. The upgraded pipeline not only improves the data extraction algorithm but also introduces several new features for users. It now enables the reduction of spectra from all available slits for both LR and XD modes. The upgraded version also implements a template-matching method for more precise wavelength calibration. Additionally, a step for flux calibration is also included. Alongside pyTANSPEC, we upgraded HxRGproc, a Python package for cleaning and generating slope images from Non-Destructive Readout (NDR) frames taken with H1RG and H2RG detectors. The package performs non-linearity correction, flags saturated pixels, removes pink noise, and eliminates cosmic ray events. HxRGproc is updated to work for the H2RG detector of TANSPEC and is set up on the TANSPEC server, ensuring users receive data that are pre-cleaned and non-linearity corrected.

astro-ph.IM↗

arXiv2026

Coarsening dynamics of fingerprint labyrinthine patterns: Machine learning assisted characterization

Fingerprint labyrinthine patterns exhibit a level of structural complexity beyond simple stripe phases, combining local stripe order with a dense network of point-like defects. Unlike symmetry-breaking phases, where coarsening proceeds via diffusive defect annihilation, or conventional stripe phases, where curvature-driven motion of extended grain boundaries dominates, the coarsening of fingerprint labyrinths is governed primarily by localized junction and terminal defects. Using the Turing-Swift-Hohenberg equation, we study the nonequilibrium relaxation of fingerprint labyrinthine patterns following a quench. To go beyond conventional Fourier-based diagnostics, we employ a template-matching convolutional neural network (TM-CNN) to identify and track junctions and terminals directly in real space, enabling a quantitative characterization of defect statistics and spatial correlations. We show that, although these point-like defects drive coarsening, their motion is strongly constrained by the surrounding stripe geometry, leading to slow, nondiffusive dynamics that are qualitatively distinct from both conventional phase ordering and stripe coarsening. Together, these results establish defect-mediated dynamics as the central organizing principle of fingerprint labyrinthine coarsening and demonstrate the effectiveness of machine-learning-assisted approaches for complex pattern-forming systems.

cond-mat.soft↗

arXiv2026

Spectral Topology and Delocalization in Disordered Hatano-Nelson Chains

The unidirectional Hatano-Nelson chain serves as the fundamental non-Hermitian building block of the Su-Schrieffer-Heeger (SSH) model. We investigate its Anderson localization properties under diagonal binary disorder. For weak disorder, the complex eigenvalue spectrum forms a single closed loop, which bifurcates into two distinct loops at a critical disorder threshold. Correspondingly, the spectral winding number {\nu} undergoes a transition from {\nu} = 1 in the weak-disorder regime, through {\nu} = 1/2 at the critical point, to {\nu} = 0 in the strong-disorder limit. We show that the eigenstates are subexponentially localized, with a localization length that varies analytically as a function of the momentum-like quantum number q. Notably, at weak and critical disorder, the spectrum hosts two completely delocalized states with diverging localization lengths. This divergence is directly correlated with the non-trivial spectral winding number. These findings remain robust under various boundary conditions, with the exception of strictly open boundaries.

cond-mat.dis-nn↗

arXiv2025

Use of solid fused silica etalon with broadband metallic coatings for calibration of high-resolution optical spectrograph

Wavelength calibration is a key factor for high-resolution spectroscopic measurements for precision radial velocities. Hollow-cathode lamps (e.g., ThAr), absorption cells (e.g., iodine cell), dielectric coated Fabry-P\'erot etalons and laser frequency combs have been implemented over the years for precise wavelength calibration and wavelength drift measurements. However, due to their various impediments as wavelength calibrators, investigations of alternative methods remain of prime interest. In this paper, we examined the feasibility of low-cost (~ $1000) commercially available solid fused silica etalon with a broadband metallic coating as a calibrator. We studied the behaviour for two cavity spacings (free spectral range of 1/cm and 0.5/cm) with temperature from theoretical derivation and experimental data. Our setup had a temperature stability of 0.8 mK for a calibrator system using an off-the-shelf dewar flask with active stabilisation. Our result from radial velocity drift measurements demonstrated that such a calibration system is capable of providing higher signal-to-noise calibration and better nightly drift measurement relative to ThAr in the wavelength range between 470 nm and 780 nm. A similar result has been previously found for Fabry-P\'erot etalons, and although the metalon solution lacks the efficiency of an etalon, it does offers a cost-effective broadband solution, which should be less prone to aging relative to complex dielectric mirror coatings. Nonetheless, long-term monitoring is required to understand the metalon behaviour in detail.

astro-ph.IM↗

arXiv2025

eARCO: Efficient Automated Root Cause Analysis with Prompt Optimization

Root cause analysis (RCA) for incidents in large-scale cloud systems is a complex, knowledge-intensive task that often requires significant manual effort from on-call engineers (OCEs). Improving RCA is vital for accelerating the incident resolution process and reducing service downtime and manual efforts. Recent advancements in Large-Language Models (LLMs) have proven to be effective in solving different stages of the incident management lifecycle including RCA. However, existing LLM-based RCA recommendations typically leverage default finetuning or retrieval augmented generation (RAG) methods with static, manually designed prompts, which lead to sub-optimal recommendations. In this work, we leverage 'PromptWizard', a state-of-the-art prompt optimization technique, to automatically identify the best optimized prompt instruction that is combined with semantically similar historical examples for querying underlying LLMs during inference. Moreover, by utilizing more than 180K historical incident data from Microsoft, we developed cost-effective finetuned small language models (SLMs) for RCA recommendation generation and demonstrate the power of prompt optimization on such domain-adapted models. Our extensive experimental results show that prompt optimization can improve the accuracy of RCA recommendations by 21% and 13% on 3K test incidents over RAG-based LLMs and finetuned SLMs, respectively. Lastly, our human evaluation with incident owners have demonstrated the efficacy of prompt optimization on RCA recommendation tasks. These findings underscore the advantages of incorporating prompt optimization into AI for Operations (AIOps) systems, delivering substantial gains without increasing computational overhead.

cs.SE↗

arXiv2025

Edge states and persistent current in a PT-symmetric extended Su-Schrieffer-Heeger model with generic boundary conditions

We consider a generalization of the Su-Schrieffer-Heeger(SSH) model by including next-nearest neighbour(NNN) interaction and balanced loss-gain(BLG), and subjecting the whole system to an external uniform magnetic field. We study the band structure, edge states and persistent current in this extended SSH model under General Boundary Condition(GBC) of which the periodic, anti-periodic and open boundary conditions appear as special cases. It is shown that the point bandgap decreases with the increasing value of the strength of the NNN interaction and vanish beyond a critical value. Further, the line gap exhibits closed-loop like structures for non-vanishing NNN interaction under the Periodic Boundary Condition(PBC). The Zak phase receives no contribution from the NNN interaction under the PBC. We show that the NNN interaction has no effect on the persistent current in the half-filled limit for the case of PBC. We show that the model without the NNN interaction is exactly solvable for a class of GBC of which PBC, anti-periodic boundary condition(APBC) and anti-hermitian boundary condition(AHBC) arise as special cases. We obtain analytic expressions for the edge states in the case of Open Boundary Condition(OBC) and AHBC for vanishing NNN interaction. We show numerically for OBC that edge states in the topologically trivial phase appear for non-vanishing NNN interaction in the parametric regions where PT-symmetry is broken under PBC. In the topologically non-trivial phase, the edge states under OBC exists only up to a critical value of the NNN strength and vanishes beyond a critical value. The bulk-boundary correspondence(BBC) for unbroken PT-phase is similar to hermitian SSH model, while non-Hermitian skin effect(NHSE) is observed for broken PT-phase.

cond-mat.mes-hall↗

arXiv2025

Orientation selection in alloy dendritic evolution during melt-pool solidification

Investigations of directionally solidifying melt pools during metal additive manufacturing (AM) reveal that the resulting subgrain cellular structures often grow along crystalline orientations different from the temperature gradient direction, some of which are not even along preferred crystallographic directions. It is well-known that dendrite orientation results from the growth competition between the heat flow direction and preferred crystallographic orientation. Specifically, the competition between interfacial anisotropy and process anisotropy (thermal gradient and growth velocity) during directional solidification leads to rich morphological diversity of the resulting dendritic structures, including tilted dendrites and seaweed patterns. The orientation selection mechanisms of such patterns remain unexplored at high velocity in the frame of AM. This study examines the tilted growth of cellular-dendritic arrays as a function of the misorientation angle ($\theta_R$) between the thermal gradient and crystal lattice directions and other relevant control parameters. We use a phase-field model to explore dendritic evolution in a binary alloy during high-velocity solidification in 2D. We find marked effects of thermal gradient, growth velocity, alloy composition, and anisotropy parameters on the possible growth directions and the primary arm spacing, constitutional undercooling, microsegregation, and secondary phases that arise during dendritic solidification. Our work provides a detailed yet concise presentation on the tilted growth and morphological transition for a broad range of thermal conditions in the high-velocity regime and the full range of $\theta_R$ for establishing orientation selection maps. These results should have qualitative relevance for controlling the subgrain structure, chemical segregation, and texture randomization in commercial dendrite materials relevant to AM.

cond-mat.soft↗

arXiv2024

Effects of isotherm patterns on cellular interface morphologies of melt pool origin

Spatiotemporal variation of the thermal gradient in the melt pool inherited from different heat input patterns or other non-equilibrium transient effects during additive manufacturing can significantly affect the resulting subgrain microstructure evolution. To examine the impact of this variation, we approximate the thermal gradient by various isotherm patterns that move with constant velocity following directional solidification. We report the first three-dimensional phase-field simulations to investigate the effects of isotherm patterns on the cellular structures typically observed in solidified melt pools. Results indicate that small variations in the isotherm can considerably impact the microstructural features. We use appropriate statistical characterizations of the solid fraction, solid percolation, and solute partitioning behavior to demonstrate the influence of isotherm patterns on the dendritic structures and semisolid mushy zones. Consistent with experimental observations, we find that non-planar isotherms produce finer cells and reduced microsegregation compared to planar isotherms. Also, we note that a tilt of the isotherm leads to a tilted state of the resulting cellular arrays. Our findings will help in understanding the qualitative aspects of the influence of temperature gradient patterns on the evolution of solidification morphologies, mushy zones, and secondary phases, which are crucial for the macroscopic description of the solidified material.

cond-mat.mtrl-sci↗

arXiv2024

Streetwise Agents: Empowering Offline RL Policies to Outsmart Exogenous Stochastic Disturbances in RTC

The difficulty of exploring and training online on real production systems limits the scope of real-time online data/feedback-driven decision making. The most feasible approach is to adopt offline reinforcement learning from limited trajectory samples. However, after deployment, such policies fail due to exogenous factors that temporarily or permanently disturb/alter the transition distribution of the assumed decision process structure induced by offline samples. This results in critical policy failures and generalization errors in sensitive domains like Real-Time Communication (RTC). We solve this crucial problem of identifying robust actions in presence of domain shifts due to unseen exogenous stochastic factors in the wild. As it is impossible to learn generalized offline policies within the support of offline data that are robust to these unseen exogenous disturbances, we propose a novel post-deployment shaping of policies (Streetwise), conditioned on real-time characterization of out-of-distribution sub-spaces. This leads to robust actions in bandwidth estimation (BWE) of network bottlenecks in RTC and in standard benchmarks. Our extensive experimental results on BWE and other standard offline RL benchmark environments demonstrate a significant improvement ($\approx$ 18% on some scenarios) in final returns wrt. end-user metrics over state-of-the-art baselines.

cs.LG↗

arXiv2024

CARMO: Dynamic Criteria Generation for Context-Aware Reward Modelling

Reward modeling in large language models is susceptible to reward hacking, causing models to latch onto superficial features such as the tendency to generate lists or unnecessarily long responses. In reinforcement learning from human feedback (RLHF) and more generally during post-training flawed reward signals often lead to outputs that optimize for these spurious correlates instead of genuine quality or correctness. We propose Context-Aware Reward Modeling (CARMO), a novel approach that first generates dynamic, context-relevant criteria to ground the reward model before producing reward scores. Unlike prior methods that rely on static rubrics, CARMO leverages large language models (LLMs) to adaptively create evaluation criteria such as logical consistency, clarity, and depth tailored to the user query. Our theoretical analysis shows that such criteria generation can mitigate reward hacking. We further demonstrate that CARMO can be distilled into smaller models, reducing the computational cost of alignment. We establish a new state-of-the-art performance in zero-shot settings for generative models, achieving a 2.1\% improvement on Reward Bench. Furthermore, alignment performed on the CARMO-curated preference dataset achieves 22.5\% and 21.1\% LC-WR and WR, respectively, on Mistral-Base (7B).

cs.CL↗

arXiv2024

TACO-RL: Task Aware Prompt Compression Optimization with Reinforcement Learning

The increasing prevalence of large language models (LLMs) such as GPT-4 in various applications has led to a surge in the size of prompts required for optimal performance, leading to challenges in computational efficiency. Prompt compression aims to reduce the inference cost by minimizing input tokens without compromising on the task performance. However, existing prompt compression techniques either rely on sub-optimal metrics such as information entropy or model it as a task-agnostic token classification problem that fails to capture task-specific information. To address these issues, we propose a novel and efficient reinforcement learning (RL) based task-aware prompt compression method. To ensure low latency requirements, we leverage existing Transformer encoder-based token classification model while guiding the learning process with task-specific reward signals using lightweight REINFORCE algorithm. We evaluate the performance of our method on three diverse and challenging tasks including text summarization, question answering and code summarization. We demonstrate that our RL-guided compression method improves the task performance by 8% - 189% across these three scenarios over state-of-the-art compression techniques while satisfying the same compression rate and latency requirements.

cs.CL↗

arXiv2024

Time series forecasting of multiphase microstructure evolution using deep learning

Microstructure evolution, which plays a critical role in determining materials properties, is commonly simulated by the high-fidelity but computationally expensive phase-field method. To address this, we approximate microstructure evolution as a time series forecasting problem within the domain of deep learning. Our approach involves implementing a cost-effective surrogate model that accurately predicts the spatiotemporal evolution of microstructures, taking an example of spinodal decomposition in binary and ternary mixtures. Our surrogate model combines a convolutional autoencoder to reduce the dimensional representation of these microstructures with convolutional recurrent neural networks to forecast their temporal evolution. We use different variants of recurrent neural networks to compare their efficacy in developing surrogate models for phase-field predictions. On average, our deep learning framework demonstrates excellent accuracy and speedup relative to the "ground truth" phase-field simulations. We use quantitative measures to demonstrate how surrogate model predictions can effectively replace the phase-field timesteps without compromising accuracy in predicting the long-term evolution trajectory. Additionally, by emulating a transfer learning approach, our framework performs satisfactorily in predicting new microstructures resulting from alloy composition and physics unknown to the model. Therefore, our approach offers a useful data-driven alternative and accelerator to the materials microstructure simulation workflow.

cond-mat.mtrl-sci↗

arXiv2024

Exploring large language models for microstructure evolution in materials

There is a significant potential for coding skills to transition fully to natural language in the future. In this context, large language models (LLMs) have shown impressive natural language processing abilities to generate sophisticated computer code for research tasks in various domains. We report the first study on the applicability of LLMs to perform computer experiments on microstructure pattern formation in model materials. In particular, we exploit LLM's ability to generate code for solving various types of phase-field-based partial differential equations (PDEs) that integrate additional physics to model material microstructures. The results indicate that LLMs have a remarkable capacity to generate multi-physics code and can effectively deal with materials microstructure problems up to a certain complexity. However, for complex multi-physics coupled PDEs for which a detailed understanding of the problem is required, LLMs fail to perform the task efficiently, since much more detailed instructions with many iterations of the same query are required to generate the desired output. Nonetheless, at their current stage of development and potential future advancements, LLMs offer a promising outlook for accelerating materials education and research by supporting beginners and experts in their physics-based methodology. We hope this paper will spur further interest to leverage LLMs as a supporting tool in the integrated computational materials engineering (ICME) approach to materials modeling and design.

cond-mat.mtrl-sci↗