arXiv subjects

Akash Gupta

Publications and source records attributed to Akash Gupta.

At least 19 recordsLinked to original sources

arXiv2026

Mechanical detection of sub-band mobilities of two-dimensional electron gas on reduced SrTiO$_3$(001) surface

The two-dimensional electron gas (2DEG) in reduced strontium titanate offers a versatile platform for oxide electronics, yet its dissipation mechanisms under field driven charge fluctuations remain poorly understood. Here, we combine low-temperature atomic force microscopy with scanning tunnelling spectroscopy to probe the force and dissipation responses of a mechanical oscillator interacting with the STO 2DEG. The observation of Rydberg like image potential states by tunnelling experiments confirm the 2DEG formation, while dissipation spectroscopy reveals bias-dependent peaks linked to local electrostatic gating and charge redistribution within the 2DEG energy sub-bands. These features are quantitatively explained by variations in quantum capacitance as carrier density is tuned by electric fields. Under magnetic fields, dissipation peaks obey the Kohler's rule, allowing extraction of carrier mobilities in each sub-band. Our results establish a non-invasive AFM - based methodology for quantifying energy losses in quantum oxides, providing new insights into charge dynamics relevant for spintronic applications.

cond-mat.mtrl-sci↗

arXiv2026

Energy Efficiency Testing and Modeling of a Commercial O-RAN System

Network energy efficiency is of critical importance to mobile network operators for economic and ecological reasons. The advent of the O-RAN architecture has brought disaggregation and virtualization, and in order to achieve the highest energy savings gains, we need rigorous measurement, analysis, and modeling of energy consumption at both the component and system levels. However, there remains a lack of publicly-available, quantitative data characterizing the behavior of commercial-grade O-RAN systems. In this white paper, we present a detailed energy-efficiency characterization and modeling of a commercial O-RAN system based on comprehensive power and performance measurements, using a network deployment that faithfully replicates a production O-RAN network deployed by a wireless carrier. The results are drawn from an energy test campaign conducted through a joint collaboration between the Open RAN Center for Integration and Deployment (ORCID) Lab Testing and Evaluation (T&E) Project and the Open Networking Foundation / Rutgers WINLAB Energy Efficiency R&D project. The test environment includes an O-RAN system with an AWS-hosted O-CU, a dedicated-server O-DU, and six high-power, multi-band O-RUs. Our results identify the dominant factors influencing power consumption across the O-RAN stack and quantify energy usage variation under different operational and traffic scenarios. These measurements can be used by operators to parameterize power-consumption models, ultimately supporting data-driven energy optimization and more sustainable operation of commercial O-RAN networks.

cs.NI↗

arXiv2025

Sub-Neptune Memories I: Implications of Inefficient Mantle Cooling and Silicate Rain

We explore the evolution of sub-Neptune (radii between $\sim$1.5 and 4 R$_\oplus$) exoplanet interior structures using our upgraded evolution code, \texttt{APPLE}, which self-consistently couples the thermal and compositional evolution of the whole structure. We incorporate stably stratified regions with convective mixing and, for the first time, ab initio results on the phase separation of silicate-hydrogen mixtures to model silicate rain in sub-Neptune envelopes. We demonstrate that inefficient mantle cooling can retain sufficient heat to Gyr ages: inefficient heat transport from mantle to envelope alone keeps radii $\sim$10\% larger than predicted by adiabatic models at late times. Silicate rain can contribute an additional $\sim$5\% to the radius, depending on envelope mass and initial metal abundance. The silicate-hydrogen immiscibility region may lie in the middle or even upper envelope, far above the envelope-mantle boundary layer, and bifurcates the envelope into two an upper, hydrogen-rich region and a lower, metal-rich region above the mantle. If silicate rain occurs, atmospheres should appear depleted of silicates while radii remain inflated at late ages. To demonstrate the effects of inefficient mantle cooling, we present interior evolution models for GJ 1214 b, K2-18 b, TOI-270 d, and TOI-1801 b, showing that hot, liquid silicate mantles with thin envelopes reproduce their radii and mean densities, providing an alternative to water-world interpretations. These results imply that bulk compositions inferred from mean density must account for the mantle thermal state and the envelope mixing/phase-separation history; such thermal ``memories'' may constrain formation entropies and temperatures when metallicities are more precisely measured.

astro-ph.EP↗

arXiv2025

The Fate of Hydrogen and Helium: From Planetary Embryos to Earth- and Neptune-like Worlds

Hydrogen, helium, silicates, and iron are key building blocks of rocky and gas-rich planets, yet their chemical interactions remain poorly constrained. Using first-principles molecular dynamics and thermodynamic integration, we quantify hydrogen and helium partitioning between molten silicate mantles and metallic cores for Earth-to-Neptune-mass planets. Hydrogen becomes strongly siderophilic above $\sim$25 GPa but weakens beyond $\sim$200 GPa, whereas helium remains lithophilic yet increasingly soluble in metal with pressure. Incorporating these trends into coupled structure-chemistry models suggests that majority of hydrogen and helium reside in planetary interiors, not atmospheres, with abundances strongly depending on planet mass. Such volatile exchange may influence the redox states of secondary atmospheres, longevity of primordial envelopes, predicted CHNOPS abundances, and emergence of helium-enriched atmospheres, while He 1083 nm and H Lyman-$\alpha$ lines provide potential probes of atmosphere-interior exchange. These findings link atomic-scale interactions to planetary-scale observables, providing new constraints on the origins of Earth-to-Neptune-sized worlds.

astro-ph.EP↗

arXiv2025

Embedding the Teacher: Distilling vLLM Preferences for Scalable Image Retrieval

Text--image retrieval is necessary for applications such as product recommendation. Embedding-based approaches like CLIP enable efficient large-scale retrieval via vector similarity search, but they are primarily trained on literal caption-like text--image pairs and often fail to capture abstract or persona-driven attributes common in product recommendation applications (e.g., ``a gift for a mother who loves gardening''). In contrast, state-of-the-art vision--language models (vLLMs) can align text with images in a flexible manner, but their limited context window prevents them from directly handling retrieval over large catalogs. We propose a framework that distills the preference rankings of a powerful vLLM into an embedding-based system, transferring its nuanced alignment abilities while maintaining the inference-time scalability of an embedding-based approach. Experiments on persona-driven product recommendation tasks demonstrate that our method significantly outperforms existing embedding-based baselines, providing an efficient solution for personalized text--image retrieval.

cs.IR↗

arXiv2025

Probing the Limits of Stylistic Alignment in Vision-Language Models

Vision-language models are increasingly used to generate image captions in specific styles, such as humor or romantic. However, these transformer-based models often struggle with this subjective task in a zero-shot setting. While preference data can be used to align them toward a desired style, such data is expensive to acquire, limiting the ability to explore the models' full capabilities. This work addresses this by studying the data efficiency of aligning small vision-language models to humor and romantic styles. This approach helps to define the performance limits of these models and determine how little preference data is needed to achieve stylistic saturation, benchmarking their capabilities and limitations.

cs.CL↗

arXiv2025

Meta-Adaptive Prompt Distillation for Few-Shot Visual Question Answering

Large Multimodal Models (LMMs) often rely on in-context learning (ICL) to perform new visual question answering (VQA) tasks with minimal supervision. However, ICL performance, especially in smaller LMMs, does not always improve monotonically when increasing the number of examples. We hypothesize that this happens because the LMM is overwhelmed by extraneous information in the image embeddings that is irrelevant to the downstream task. To address this, we propose a meta-learning approach that induces few-shot capabilities in LMMs through a fixed set of soft prompts distilled from task-relevant visual features, which are adapted at test time using a small number of examples. We facilitate this distillation through an attention-mapper module that can be easily integrated with any LMM architecture and is jointly learned with soft prompts. Evaluation on the VL-ICL Bench shows that our method successfully achieves task adaptation in low-data regimes with just a few gradient steps, outperforming ICL by 21.2%. Comparisons with parameter-efficient finetuning methods demonstrate that meta-learning further enhances this adaptation by 7.7% for various VQA tasks.

cs.AI↗

arXiv2025

Are Kronberger 80 and/or Kronberger 82 regions PeVatron candidates for LHAASO J2108+5157?

High-energy gamma rays have been detected in the region of LHAASO~J2108+5157 by the Fermi--LAT, HAWC and LHAASO-KM2A observatories. Cygnus~OB2 in Cygnus--X has been confirmed as the first strong stellar cluster PeVatron in our Galaxy. Thus, the star--forming regions Kronberger~80 and Kronberger~82, located in the field of LHAASO~J2108+5157, are analyzed to evaluate their stellar population and potential as associated PeVatron candidates. A distance of 10~kpc is adopted for Kronberger~80, while $\sim$1.6~kpc is estimated for Kronberger~82. Based on stellar densities, we report that their cluster radii are 2.5$\arcmin$ and 2.0$\arcmin$, while IR photometry reveals poor stellar content in massive O-type stars in both cases. From optical data, the estimation of cluster ages are 5--12.6~Myr and $\lesssim$ 5~Myr, respectively. We conclude that, in contrast to the stellar content of Cygnus~OB2, it is unlikely that Kronberger~80 and Kronberger~82 are PeVatrons associated with LHAASO~J2108+5157. The presence of a PeVatron in this region remains a mystery, but we confirm that the two Kronberger regions are star--forming regions undergoing formation rather than evolution.

astro-ph.HE↗

arXiv2025

Signatures of Atmospheric Mass Loss and Planet Migration in the Time Evolution of Short-Period Transiting Exoplanets

Comparative studies of young and old exoplanet populations offer a glimpse into how planets may form and evolve with time. We present an occurrence rate study of short-period ($<$12 days) planets between 1.8--10 Rearth around 1374 FGK stars in nearby (200 pc) young clusters ($<$1 Gyr), utilizing data from the Transiting Exoplanet Survey Satellite (TESS) mission. These planets represent a population closer to their primordial state. We find that the occurrence rate of young planets is higher ($64^{+32}_{-22}$%) compared to the Gyr-old population observed by \kepler ($7.98^{+0.37}_{-0.35}$%). Dividing our sample into bins of young (10--100 Myr) and intermediate (100\,Myr--1 Gyr) ages, we also find that the occurrence distribution in orbital period remains unchanged while the distribution in planet radius changes with time. Specifically, the radius distribution steepens with age, indicative of a larger planet population shrinking due to the atmospheric thermal cooling and mass loss. We also find evidence for an increase (1.9$\sigma$) in occurrence after 100 Myr, possibly due to tidal migration driving planets inside of 12 days. While evidence suggests post-disk migration and atmospheric mass loss shape the population of short-period planets, more detections of young planets are needed to improve statistical comparisons with older planets. Detecting long-period young planets and planets $<$1.8 Rearth will help us understand these processes better. Additionally, studying young planetary atmospheres provides insights into planet formation and the efficiency of atmospheric mass loss mechanisms on the evolution of planetary systems.

astro-ph.EP↗

arXiv2025

ZeroBench: An Impossible Visual Benchmark for Contemporary Large Multimodal Models

Large Multimodal Models (LMMs) exhibit shortfalls when interpreting images and, by some measures, have poorer spatial cognition than young children or animals. Despite this, they attain high scores on many popular visual benchmarks, with headroom rapidly eroded by model progress. This creates a need for difficult benchmarks that remain relevant for longer. We introduce ZeroBench - a lightweight visual reasoning benchmark curated using adversarial filtering to be "impossible" for frontier LMMs at its original release, with initial SotA scores of 0% pass@1 and pass^5. We track progress on ZeroBench over the subsequent year, observing SotA reaching 6% pass^5 and 19% pass@5, indicating the potential longevity of the benchmark. We evaluate 46 LMMs on ZeroBench, compare performance to a human baseline, analyse strengths and weaknesses, chart a year of progress in visual capabilities, and publicly release ZeroBench at https://zerobench.github.io.

cs.CV↗

arXiv2024

Obscured star clusters in the Inner Milky Way. How many massive young clusters are still awaiting detection?

Aims. Our goal is twofold. First, to detect new clusters we apply the newest methods for the detection of clustering with the best available wide-field sky surveys in the mid-infrared because they are the least affected by extinction. Second, we address the question of cluster detection's completeness, for now limiting it to the most massive star clusters. Methods. This search is based on the mid-infrared Galactic Legacy Infrared Mid Plane Survey Extraordinaire (GLIMPSE), to minimize the effect of dust extinction. The search Ordering Points To Identify the Clustering Structure (OPTICS) clustering algorithm is applied to identify clusters, after excluding the bluest, presumably foreground sources, to improve the cluster-to-field contrast. The success rate for cluster identification is estimated with a semi-empirical simulation that adds clusters, based on the real objects, to the point source catalog, to be recovered later with the same search algorithm that was used in the search for new cluster candidates. As a first step, this is limited to the most massive star clusters with a total mass of 104 $M_\odot$. Results. Our automated search, combined with inspection of the color-magnitude diagrams and images yielded 659 cluster candidates; 106 of these appear to have been previously identified, suggesting that a large hidden population of star clusters still exists in the inner Milky Way. However, the search for the simulated supermassive clusters achieves a recovery rate of 70 to 95%, depending on the distance and extinction toward them. Conclusions. The new candidates, if confirmed, indicate that the Milky Way still harbors a sizeable population of still unknown clusters. However, they must be objects of modest richness, because our simulation indicates that there is no substantial hidden population of supermassive clusters in the central region of our Galaxy.

astro-ph.GA↗

arXiv2024

The miscibility of hydrogen and water in planetary atmospheres and interiors

Many planets in the solar system and across the galaxy have hydrogen-rich atmospheres overlying more heavy element-rich interiors with which they interact for billions of years. Atmosphere-interior interactions are thus crucial to understanding the formation and evolution of these bodies. However, this understanding is still lacking in part because the relevant pressure-temperature conditions are extreme. We conduct molecular dynamics simulations based on Density Functional Theory to investigate how hydrogen and water interact over a wide range of pressure and temperature, encompassing the interiors of Neptune-sized and smaller planets. We determine the critical curve at which a single homogeneous phase exsolves into two separate, hydrogen-rich and water-rich phases, finding good agreement with existing experimental data. We find that the temperature along the critical curve increases with increasing pressure and shows the influence of a change in fluid structure from molecular to atomic near 30 GPa and 3000 K, which may impact magnetic field generation. The internal temperatures of many exoplanets, including TOI-270 d and K2-18 b may lie entirely above the critical curve: the envelope is expected to consist of a single homogeneous hydrogen-water fluid, that is much less susceptible to atmospheric loss as compared with a pure hydrogen envelope. As planets cool, they cross the critical curve, leading to rainout of water-rich fluid and an increase in internal luminosity. Compositions of the resulting outer, hydrogen-rich, and inner, water-rich envelopes depend on age and instellation and are governed by thermodynamics. Rainout of water may be occurring in Uranus and Neptune at present.

astro-ph.EP↗

arXiv2024

LLM Task Interference: An Initial Study on the Impact of Task-Switch in Conversational History

With the recent emergence of powerful instruction-tuned large language models (LLMs), various helpful conversational Artificial Intelligence (AI) systems have been deployed across many applications. When prompted by users, these AI systems successfully perform a wide range of tasks as part of a conversation. To provide some sort of memory and context, such approaches typically condition their output on the entire conversational history. Although this sensitivity to the conversational history can often lead to improved performance on subsequent tasks, we find that performance can in fact also be negatively impacted, if there is a task-switch. To the best of our knowledge, our work makes the first attempt to formalize the study of such vulnerabilities and interference of tasks in conversational LLMs caused by task-switches in the conversational history. Our experiments across 5 datasets with 15 task switches using popular LLMs reveal that many of the task-switches can lead to significant performance degradation.

cs.CL↗

arXiv2023

ModEFormer: Modality-Preserving Embedding for Audio-Video Synchronization using Transformers

Lack of audio-video synchronization is a common problem during television broadcasts and video conferencing, leading to an unsatisfactory viewing experience. A widely accepted paradigm is to create an error detection mechanism that identifies the cases when audio is leading or lagging. We propose ModEFormer, which independently extracts audio and video embeddings using modality-specific transformers. Different from the other transformer-based approaches, ModEFormer preserves the modality of the input streams which allows us to use a larger batch size with more negative audio samples for contrastive learning. Further, we propose a trade-off between the number of negative samples and number of unique samples in a batch to significantly exceed the performance of previous methods. Experimental results show that ModEFormer achieves state-of-the-art performance, 94.5% for LRS2 and 90.9% for LRS3. Finally, we demonstrate how ModEFormer can be used for offset detection for test clips.

cs.CV↗

arXiv2022

GAMA: Generative Adversarial Multi-Object Scene Attacks

The majority of methods for crafting adversarial attacks have focused on scenes with a single dominant object (e.g., images from ImageNet). On the other hand, natural scenes include multiple dominant objects that are semantically related. Thus, it is crucial to explore designing attack strategies that look beyond learning on single-object scenes or attack single-object victim classifiers. Due to their inherent property of strong transferability of perturbations to unknown models, this paper presents the first approach of using generative models for adversarial attacks on multi-object scenes. In order to represent the relationships between different objects in the input scene, we leverage upon the open-sourced pre-trained vision-language model CLIP (Contrastive Language-Image Pre-training), with the motivation to exploit the encoded semantics in the language space along with the visual space. We call this attack approach Generative Adversarial Multi-object scene Attacks (GAMA). GAMA demonstrates the utility of the CLIP model as an attacker's tool to train formidable perturbation generators for multi-object scenes. Using the joint image-text features to train the generator, we show that GAMA can craft potent transferable perturbations in order to fool victim classifiers in various attack settings. For example, GAMA triggers ~16% more misclassification than state-of-the-art generative approaches in black-box settings where both the classifier architecture and data distribution of the attacker are different from the victim. Our code is available here: https://abhishekaich27.github.io/gama.html

cs.CV↗

arXiv2022

UltraMNIST Classification: A Benchmark to Train CNNs for Very Large Images

Convolutional neural network (CNN) approaches available in the current literature are designed to work primarily with low-resolution images. When applied on very large images, challenges related to GPU memory, smaller receptive field than needed for semantic correspondence and the need to incorporate multi-scale features arise. The resolution of input images can be reduced, however, with significant loss of critical information. Based on the outlined issues, we introduce a novel research problem of training CNN models for very large images, and present 'UltraMNIST dataset', a simple yet representative benchmark dataset for this task. UltraMNIST has been designed using the popular MNIST digits with additional levels of complexity added to replicate well the challenges of real-world problems. We present two variants of the problem: 'UltraMNIST classification' and 'Budget-aware UltraMNIST classification'. The standard UltraMNIST classification benchmark is intended to facilitate the development of novel CNN training methods that make the effective use of the best available GPU resources. The budget-aware variant is intended to promote development of methods that work under constrained GPU memory. For the development of competitive solutions, we present several baseline models for the standard benchmark and its budget-aware variant. We study the effect of reducing resolution on the performance and present results for baseline models involving pretrained backbones from among the popular state-of-the-art models. Finally, with the presented benchmark dataset and the baselines, we hope to pave the ground for a new generation of CNN methods suitable for handling large images in an efficient and resource-light manner.

cs.CV↗

arXiv2022

Poisson2Sparse: Self-Supervised Poisson Denoising From a Single Image

Image enhancement approaches often assume that the noise is signal independent, and approximate the degradation model as zero-mean additive Gaussian. However, this assumption does not hold for biomedical imaging systems where sensor-based sources of noise are proportional to signal strengths, and the noise is better represented as a Poisson process. In this work, we explore a sparsity and dictionary learning-based approach and present a novel self-supervised learning method for single-image denoising where the noise is approximated as a Poisson process, requiring no clean ground-truth data. Specifically, we approximate traditional iterative optimization algorithms for image denoising with a recurrent neural network that enforces sparsity with respect to the weights of the network. Since the sparse representations are based on the underlying image, it is able to suppress the spurious components (noise) in the image patches, thereby introducing implicit regularization for denoising tasks through the network structure. Experiments on two bio-imaging datasets demonstrate that our method outperforms the state-of-the-art approaches in terms of PSNR and SSIM. Our qualitative results demonstrate that, in addition to higher performance on standard quantitative metrics, we are able to recover much more subtle details than other compared approaches. Our code is made publicly available at https://github.com/tacalvin/Poisson2Sparse

eess.IV↗

arXiv2022

Properties of the radius valley around low mass stars: Predictions from the core-powered mass-loss mechanism

In recent years, analyzing the bimodality in the size distribution of small planets, i.e., the `radius valley', has given us unprecedented insight into the planet formation process. Here we explore the properties of the radius valley for low mass stars, assuming that the core-powered mass-loss is the dominant process shaping the small exoplanet population. We show that the slope of radius valley in the planet size-orbital period space, to first-order, does not vary with stellar mass and has a negative slope of $\text{d log}R_p/\text{d log}P \simeq -0.11$ even for stars as small as 0.1 $M_\odot$, as observed in latest studies. Furthermore, we find that the slope of the radius valley in the planet size-stellar mass space is $\text{d log}R_p/\text{d log}M_\ast \simeq (3 \zeta - 2)/36$ where $\zeta$ is given by the stellar mass-luminosity relation $L_\ast \propto M_\ast^\zeta$. Because $\zeta$ is $\gtrsim$ 2 and increases with stellar mass, we predict that the radius valley has a positive slope in the planet size-stellar mass space across FGKM dwarfs. This slope, however, decreases (increases) in magnitude towards lower (higher) mass stars, due to the variation of $\zeta$ with stellar mass. While around 1.0 $M_\odot$ stars the slope is $\text{d log}R_p/\text{d log}M_\ast \sim 0.37$, it is as low as $\sim 0.13$ around 0.1 $M_\odot$ stars. In addition, we find that the radius valley is narrower and less empty around lower mass stars. Finally, we show that predictions for the radius valley for core-powered mass-loss and photoevaporation become increasingly distinct for lower mass stars.

astro-ph.EP↗