arXiv subjects

Ke Zhang

Publications and source records attributed to Ke Zhang.

At least 109 records · Page 6Linked to original sources

arXiv2025

HydraInfer: Hybrid Disaggregated Scheduling for Multimodal Large Language Model Serving

Multimodal Large Language Models (MLLMs) have been rapidly advancing, enabling cross-modal understanding and generation, and propelling artificial intelligence towards artificial general intelligence. However, existing MLLM inference systems are typically designed based on the architecture of language models, integrating image processing and language processing as a single scheduling unit. This design struggles to accommodate the heterogeneous demands of different stages in terms of computational resources, memory access patterns, and service-level objectives (SLOs), leading to low resource utilization and high request latency, ultimately failing to meet the service requirements of diverse inference scenarios. To address these challenges, we propose HydraInfer, an efficient MLLM inference system that adopts a Hybrid Encode-Prefill-Decode (EPD) Disaggregation architecture. By scheduling the three stages - encode, prefill, and decode - onto separate heterogeneous inference instances, the system flexibly reallocates resources across stages, significantly reducing idle computation, alleviating resource bottlenecks, and improving overall system throughput and scalability. In addition, HydraInfer supports a stage-level batching strategy that enhances load balancing, enables parallel execution of visual and language models, and further optimizes inference performance. Experiments under real multimodal inference workloads demonstrate that HydraInfer can achieve up to 4x higher inference throughput compared to state-of-the-art systems (e.g., vLLM) on a single-node 8xH800 GPU cluster, while meeting the 90th percentile request SLO.

cs.DC↗

arXiv2025

Searching for Exoplanets Born Outside the Milky Way: VOYAGERS Survey Design

Observations over the past few decades have found that planets are common around nearby stars in our Galaxy, but little is known about planets that formed outside the Milky Way. We describe the design and early implementation of a survey to test whether planets also exist orbiting the remnant stars of ancient dwarf galaxies that merged with the Milky Way, and if so, how they differ from their Milky Way counterparts. VOYAGERS (Views Of Yore - Ancient Gaia-enceladus Exoplanet Revealing Survey) is a radial velocity (RV) search using precision spectrographs to discover exoplanets orbiting very low metallicity ($-2.8 < [\mathrm{Fe/H}] \leq -0.8$) stars born in the dwarf galaxy Enceladus, which merged with the Milky Way galaxy about 10 Gyr ago. A sample of 22 candidates have been screened from a catalog of Gaia-Enceladus-Sausage (GES) members using a combination of stellar properties and reconnaissance observations from the TRES spectrograph. Precision RV measurements have been initiated using the NEID, HARPS-N, and CARMENES spectrographs. We plan to focus most upcoming observations on 10 main sequence targets. Data collection is well underway, with 778 observations on 22 candidates (385 of which are 10 focus targets), but far from complete. This survey is designed to be sensitive to sub-Neptune mass planets with periods up to hundreds of days. We note that the RV analysis gives mass multiplied by $\sin (inclination)$ or the minimum mass for exoplanets. The expected survey yield is three planets, assuming that occurrence rates are similar to those in the Milky Way and taking into account the degeneracy with inclination in our yield models. Our survey is designed to detect at least one exoplanet if occurrence rates are similar to known Milky Way exoplanets or, if no exoplanets are discovered, to rule out a Milky Way-like planet population in GES with 95% confidence level.

astro-ph.EP↗

arXiv2025

Arrow: Adaptive Scheduling Mechanisms for Disaggregated LLM Inference Architecture

Existing large language model (LLM) serving systems typically employ Prefill-Decode disaggregated architecture to prevent computational interference between the prefill and decode phases. However, in real-world LLM serving scenarios, significant fluctuations in request input/output lengths lead to imbalanced computational loads between prefill and decode nodes under traditional static node allocation strategies, consequently preventing efficient utilization of computing resources to improve the system's goodput. To address this challenge, we design and implement Arrow, an adaptive scheduler that leverages stateless instances and latency characteristics of prefill and decode tasks to achieve efficient adaptive request and instance scheduling. Arrow dynamically adjusts the number of instances handling prefill and decode tasks based on real-time cluster performance metrics, substantially enhancing the system's capability to handle traffic spikes and load variations. Our evaluation under diverse real-world workloads shows that Arrow achieves up to $2.55 \times$ higher request serving rates compared to state-of-the-art Prefill-Decode disaggregated serving systems.

cs.DC↗

arXiv2025

LIFT: Interpretable truck driving risk prediction with literature-informed fine-tuned LLMs

This study proposes an interpretable prediction framework with literature-informed fine-tuned (LIFT) LLMs for truck driving risk prediction. The framework integrates an LLM-driven Inference Core that predicts and explains truck driving risk, a Literature Processing Pipeline that filters and summarizes domain-specific literature into a literature knowledge base, and a Result Evaluator that evaluates the prediction performance as well as the interpretability of the LIFT LLM. After fine-tuning on a real-world truck driving risk dataset, the LIFT LLM achieved accurate risk prediction, outperforming benchmark models by 26.7% in recall and 10.1% in F1-score. Furthermore, guided by the literature knowledge base automatically constructed from 299 domain papers, the LIFT LLM produced variable importance ranking consistent with that derived from the benchmark model, while demonstrating robustness in interpretation results to various data sampling conditions. The LIFT LLM also identified potential risky scenarios by detecting key combination of variables in truck driving risk, which were verified by PERMANOVA tests. Finally, we demonstrated the contribution of the literature knowledge base and the fine-tuning process in the interpretability of the LIFT LLM, and discussed the potential of the LIFT LLM in data-driven knowledge discovery.

cs.AI↗

arXiv2025

Quantification of Electrolyte Degradation in Lithium-ion Batteries with Neutron Imaging Techniques

Non-destructive characterization of lithium-ion batteries provides critical insights for optimizing performance and lifespan while preserving structural integrity. Optimizing electrolyte design in commercial LIBs requires consideration of composition, electrolyte-to-capacity ratio, spatial distribution, and associated degradation pathways. However, existing non-destructive methods for studying electrolyte infiltration, distribution, and degradation in LIBs lack the spatiotemporal resolution required for precise observation and quantification of the electrolyte. In this study, we employ neutron imaging with sufficient spatial resolution ~150 um and large field of view 20x20 cm2 to quantitatively resolve the electrolyte inventory and distribution within LiFePO4/graphite pouch cells under high-temperature accelerated aging. Quantitative standard curves based on neutron transmission attenuation reveal a clear electrolyte dry-out threshold at 3.18 g Ah-1 and the two stages evolutions of EI during cell aging were quantified. By integrating non-destructive electrochemical diagnostics, accelerated graphite material loss and liquid phase Li+ diffusion degradation is observed during pore-drying. Further analysis, including operando cyclic aging, reveals that the neutron transmission below the saturation reference is due to the enrichment of hydrogen nuclei within the solid-electrolyte interphase. Assumed pore-drying does not occur, the SEI signal of the electrodes can be quantitatively decoupled during ageing. Combined analyses with NI, TOF-SIMS, and SEM reveal that high EI cells exhibit uniform SEI growth and reduced degradation, while low EI cells show uneven SEI formation, accelerating capacity loss. This study unveils a dynamic electrolyte infiltration-consumption-dry-out process in LIBs, offering non-destructive and quantitative insights to guide sustainable and durable battery development.

physics.app-ph↗

arXiv2025

Think Before You Diffuse: Infusing Physical Rules into Video Diffusion

Recent video diffusion models have demonstrated their great capability in generating visually-pleasing results, while synthesizing the correct physical effects in generated videos remains challenging. The complexity of real-world motions, interactions, and dynamics introduce great difficulties when learning physics from data. In this work, we propose DiffPhy, a generic framework that enables physically-correct and photo-realistic video generation by fine-tuning a pre-trained video diffusion model. Our method leverages large language models (LLMs) to infer rich physical context from the text prompt. To incorporate this context into the video diffusion model, we use a multimodal large language model (MLLM) to verify intermediate latent variables against the inferred physical rules, guiding the gradient updates of model accordingly. Textual output of LLM is transformed into continuous signals. We then formulate a set of training objectives that jointly ensure physical accuracy and semantic alignment with the input text. Additionally, failure facts of physical phenomena are corrected via attention injection. We also establish a high-quality physical video dataset containing diverse phyiscal actions and events to facilitate effective finetuning. Extensive experiments on public benchmarks demonstrate that DiffPhy is able to produce state-of-the-art results across diverse physics-related scenarios. Our project page is available at https://bwgzk-keke.github.io/DiffPhy/.

cs.CV↗

arXiv2025

FreeViS: Training-free Video Stylization with Inconsistent References

Video stylization plays a key role in content creation, but it remains a challenging problem. Naïvely applying image stylization frame-by-frame hurts temporal consistency and reduces style richness. Alternatively, training a dedicated video stylization model typically requires paired video data and is computationally expensive. In this paper, we propose FreeViS, a training-free video stylization framework that generates stylized videos with rich style details and strong temporal coherence. Our method integrates multiple stylized references to a pretrained image-to-video (I2V) model, effectively mitigating the propagation errors observed in prior works, without introducing flickers and stutters. In addition, it leverages high-frequency compensation to constrain the content layout and motion, together with flow-based motion cues to preserve style textures in low-saliency regions. Through extensive evaluations, FreeViS delivers higher stylization fidelity and superior temporal consistency, outperforming recent baselines and achieving strong human preference. Our training-free pipeline offers a practical and economic solution for high-quality, temporally coherent video stylization. The code and videos can be accessed via https://xujiacong.github.io/FreeViS/

cs.CV↗

arXiv2025

Observational Constraints on Evolution of Dust Disc Properties in Upper Scorpius

Protoplanetary discs in the Upper Scorpius star-forming region are excellent laboratories to investigate late stages of planet formation. In this work, we analyse the morphology of the dust continuum emission of 121 discs from an ALMA Band 7 survey of the Upper Scorpius region. This analysis is done in the visibility plane, to measure the flux, geometry and characterise potential structures. We compare the results with state-of-the art gas and dust evolution models that include external photoevaporation, with mild values of the $F_{\rm{UV}}$ of 1-40$G_0$. From the visibility analysis, 52 of the 121 discs are resolved (43%). From the resolved discs, 24 discs have structures and 28 remain as smooth discs at the mean resolution scale of $\sim$0.1$^{\prime \prime}$ (~14au). Our results show no significant dust disc size evolution of the surviving discs in UpperSco when compared to discs in younger star-forming regions, such as Lupus. We find a strong, steeper-than-previously-reported correlation between dust disc size and disc millimeter continuum luminosity, in agreement with drift-dominated dust evolution models. We also find positive correlations between the dust disc mass vs. stellar mass and dust disc size vs. stellar mass. The slope of the dust disc size vs. stellar mass relationship is steeper compared to younger star forming regions. Additionally, we observe no significant correlation between dust disc properties and the environmental $F_{\rm{UV}}$, consistent with models predicting that dust disc properties are primarily shaped by drift and dust traps. Our models predict that gas disc masses and sizes should be highly affected by the moderate $F_{\rm{UV}}$ values that Upper Scorpius discs experience in contrast to the dust, highlighting the need for deeper and higher-resolution gas observations of these discs exposed to mild external photoevaporation.

astro-ph.EP↗

arXiv2025

Symphony: A Decentralized Multi-Agent Framework for Scalable Collective Intelligence

Most existing Large Language Model (LLM)-based agent frameworks rely on centralized orchestration, incurring high deployment costs, rigid communication topologies, and limited adaptability. To address these challenges, we introduce Symphony, a decentralized multi-agent system which enables lightweight LLMs on consumer-grade GPUs to coordinate. Symphony introduces three key mechanisms: (1) a decentralized ledger that records capabilities, (2) a Beacon-selection protocol for dynamic task allocation, and (3) weighted result voting based on CoTs. This design forms a privacy-saving, scalable, and fault-tolerant orchestration with low overhead. Empirically, Symphony outperforms existing baselines on reasoning benchmarks, achieving substantial accuracy gains and demonstrating robustness across models of varying capacities.

cs.LG↗

arXiv2025

OpenM3D: Open Vocabulary Multi-view Indoor 3D Object Detection without Human Annotations

Open-vocabulary (OV) 3D object detection is an emerging field, yet its exploration through image-based methods remains limited compared to 3D point cloud-based methods. We introduce OpenM3D, a novel open-vocabulary multi-view indoor 3D object detector trained without human annotations. In particular, OpenM3D is a single-stage detector adapting the 2D-induced voxel features from the ImGeoNet model. To support OV, it is jointly trained with a class-agnostic 3D localization loss requiring high-quality 3D pseudo boxes and a voxel-semantic alignment loss requiring diverse pre-trained CLIP features. We follow the training setting of OV-3DET where posed RGB-D images are given but no human annotations of 3D boxes or classes are available. We propose a 3D Pseudo Box Generation method using a graph embedding technique that combines 2D segments into coherent 3D structures. Our pseudo-boxes achieve higher precision and recall than other methods, including the method proposed in OV-3DET. We further sample diverse CLIP features from 2D segments associated with each coherent 3D structure to align with the corresponding voxel feature. The key to training a highly accurate single-stage detector requires both losses to be learned toward high-quality targets. At inference, OpenM3D, a highly efficient detector, requires only multi-view images for input and demonstrates superior accuracy and speed (0.3 sec. per scene) on ScanNet200 and ARKitScenes indoor benchmarks compared to existing methods. We outperform a strong two-stage method that leverages our class-agnostic detector with a ViT CLIP-based OV classifier and a baseline incorporating multi-view depth estimator on both accuracy and speed.

cs.CV↗

arXiv2025

Interpolating Speaker Identities in Embedding Space for Data Expansion

The success of deep learning-based speaker verification systems is largely attributed to access to large-scale and diverse speaker identity data. However, collecting data from more identities is expensive, challenging, and often limited by privacy concerns. To address this limitation, we propose INSIDE (Interpolating Speaker Identities in Embedding Space), a novel data expansion method that synthesizes new speaker identities by interpolating between existing speaker embeddings. Specifically, we select pairs of nearby speaker embeddings from a pretrained speaker embedding space and compute intermediate embeddings using spherical linear interpolation. These interpolated embeddings are then fed to a text-to-speech system to generate corresponding speech waveforms. The resulting data is combined with the original dataset to train downstream models. Experiments show that models trained with INSIDE-expanded data outperform those trained only on real data, achieving 3.06\% to 5.24\% relative improvements. While INSIDE is primarily designed for speaker verification, we also validate its effectiveness on gender classification, where it yields a 13.44\% relative improvement. Moreover, INSIDE is compatible with other augmentation techniques and can serve as a flexible, scalable addition to existing training pipelines.

eess.AS↗

arXiv2025

The ALMA Survey of Gas Evolution of PROtoplanetary Disks (AGE-PRO): II. Dust and Gas Disk Properties in the Ophiuchus Star-forming Region

The ALMA survey of Gas Evolution in PROtoplanetary disks (AGE-PRO) Large Program aims to trace the evolution of gas disk mass and size throughout the lifetime of protoplanetary disks. This paper presents Band-6 ALMA observations of 10 embedded (Class I and Flat Spectrum) sources in the Ophiuchus molecular cloud, with spectral types ranging from M3 to K6 stars, which serve as the evolutionary starting point in the AGE-PRO sample. While we find 4 nearly edge on disks (>70 deg.), and 3 highly inclined disks (>60 deg.) in our sample, we show that, as a population, embedded disks in Ophiuchus are not significantly contaminated by more evolved, but highly inclined sources. We derived dust disk masses from the Band 6 continuum and estimated gas disk masses from the C18O and C17O lines. The mass estimates from the C17O line are slightly higher, suggesting C18O emission might be partially optically thick. While the 12CO and 13CO lines are severely contaminated by extended emission and self-absorption, the C18O and C17O lines allowed us to trace the radial extent of the gaseous disks. From these measurements, we found that the C18O and C17O fluxes correlate well with each other and with the continuum fluxes. Furthermore, the C18O and C17O lines present a larger radial extension than disk dust sizes by factors ranging from 1.5 to 2.5, as it is found for Class II disks using the radial extension of the 12CO. In addition, we have detected outflows in three disks from 12CO observations.

astro-ph.EP↗

arXiv2025

The ALMA Survey of Gas Evolution of PROtoplanetary Disks (AGE-PRO): VI. Comparison of Dust Evolution Models to AGE-PRO Observations

The potential for planet formation of a circumstellar disk depends on the dust and gas reservoirs, which evolve as a function of the disk age. The ALMA Large Program AGE-PRO has measured several disk properties across three star-forming regions of different ages, and in this study we compare the observational results to dust evolution simulations. Using DustPy for the dust evolution, and RADMC-3D for the radiative transfer, we ran a large grid of models spanning stellar masses of 0.25, 0.50, 0.75, and 1.0 $M_\odot$, with different initial conditions, including: disk sizes, disk gas masses, and dust-to-gas ratio, and viscosity. Our models are performed assuming smooth, weakly, or strongly substructured disks, aiming to investigate if any observational trend can favor or exclude the presence of dust traps. The observed gas masses in the disks of the AGE-PRO sample are not reproducible with our models, which only consider viscous evolution with constant $α$, suggesting that additional physical mechanisms play a role in the evolution of the gas mass of disks. When comparing the dust continuum emission fluxes and sizes at 1.3 mm, we find that most of the disks in the AGE-PRO sample are consistent with simulations that have either weak or strong dust traps. The evolution of spectral index in the AGE-PRO sample is also suggestive of an unresolved population of dust traps. Future observations at high angular resolution are still needed to test several hypotheses that result from comparing the observations to our simulations, including that more massive disks in gas mass have the potential to form dust traps at larger disk radii.

astro-ph.EP↗

arXiv2025

The ALMA Survey of Gas Evolution of PROtoplanetary Disks (AGE-PRO): I. Program Overview and Summary of First Results

We present the ALMA Survey of Gas Evolution of PROtoplanetary Disks (AGE-PRO), a Large Program of the Atacama Large Millimeter/submillimeter Array (ALMA). AGE-PRO aims to systematically trace the evolution of gas disk mass and size throughout the lifetime of protoplanetary disks. It uses a carefully selected sample of 30 disks around M3-K6 stars in three nearby star-forming regions: Ophiuchus (0.5-1 Myr), Lupus (1-3 Myr), and Upper Sco (2-6 Myr). Assuming the three regions had similar initial conditions and evolutionary paths, we find the median gas disk mass appears to decrease with age. Ophiuchus disks have the highest median gas mass (6 M$_{\rm Jup}$), while the Lupus and Upper Sco disks have significantly lower median masses (0.68 and 0.44 M$_{\rm Jup}$, respectively). Notably, the gas and dust disk masses appear to evolve on different timescales. This is evidenced by the median gas-to-dust mass ratio, which decreases from 122 in the youngest disks ($<$1 Myr) to 46 in Lupus disks, and then increases to 120 in the Upper Sco disks. The median gas disk sizes range between 74-110 au, suggesting that typical gas disks are much smaller than those of well-studied, massive disks. Population synthesis models suggest that magneto-hydrodynamic wind-driven accretion can reproduce median disk properties across all three regions, when assuming compact disks with a declining magnetic field over time. In contrast, turbulent-driven models overestimate gas masses of $>$1 Myr disks by an order of magnitude. Here we discuss the program's motivation, survey design, sample selection, observation and data calibration processes, and highlight the initial results.

astro-ph.EP↗

arXiv2025

Cosmic cascades: How disk substructure regulates the flow of water to inner planetary systems

The influx of icy pebbles to the inner regions of protoplanetary disks constitutes a fundamental ingredient in most planet formation theories. The observational determination of the magnitude of this pebble flux and its dependence on disk substructure (disk gaps as pebble traps) would be a significant step forward. In this work we analyze a sample of 21 T Tauri disks (with ages $\approx 0.5{-}2\mathrm{~Myr}$) using JWST/MIRI spectra homogeneously reduced with the JDISCS pipeline and high-angular-resolution ALMA continuum data. We find that the 1500/6000 K water line flux ratio measured with JWST - a tracer of cold water vapor and pebble drift near the snowline - correlates with the radial location of the innermost dust gap in ALMA continuum observations (ranging from 8.7 to 69 au), confirming predictions from recent models that study connections between the inner and outer disk reservoirs. We develop a population synthesis exploration of pebble drift in gapped disks and find a good match to the observed trend for early and relatively effective gaps, while scenarios where pebble drift happens quickly, gaps are very leaky, or where gaps form late are disfavored on a population level. Inferred snowline pebble mass fluxes (ranging between $10^{-6}$ and $10^{-3}~M_\oplus/\mathrm{yr}$ depending on gap position) are comparable to fluxes used in pebble accretion studies and those proposed for the inner Solar System, while system-to-system variations suggest differences in the emerging planetary system architectures and water budgets.

astro-ph.EP↗

arXiv2025

The ALMA Survey of Gas Evolution of PROtoplanetary Disks (AGE-PRO): VIII. The impact of external photoevaporation on disk masses and radii in Upper Scorpius

Protoplanetary disk evolution can be deeply influenced by the UV radiation emitted by neighboring massive stars (mainly of spectral type O and B). We show that the process of external photoevaporation, which causes an outside-in depletion of disk material due to environmental UV radiation, can lead to a significant decrease in disk size, and moderate in disk mass and lifetime even at moderate irradiation levels (1-10 G$_{0}$). In this work we investigate the role of external photoevaporation in shaping the masses and sizes of the ten AGE-PRO disks in the Upper Scorpius region, which we estimate to be subject to FUV fluxes ranging between 2 and 12 G$_{0}$, on average. We compare the disk masses and sizes resulting from 1D numerical viscous evolution simulations in which the effect of external photoevaporation is included, to the values retrieved from the AGE-PRO observations. While the pure viscous framework fails in adequately explaining the observed disk properties in Upper Scorpius, with the inclusion of external photoevaporation we can successfully reproduce gas disk sizes for 7 out of 10 sources within a factor <2, when the initial disk mass is 1-10% of the stellar mass. We emphasize the importance of accounting for the environmental irradiation when comparing star-forming regions of different ages, even when moderate FUV irradiation fields are experienced, as in the case of Upper Scorpius.

astro-ph.EP↗

arXiv2025

The ALMA Survey of Gas Evolution of PROtoplanetary Disks (AGE-PRO): IV. Dust and Gas Disk Properties in the Upper Scorpius Star-forming Region

The Atacama Large Millimeter/submillimeter Array (ALMA) large program AGE-PRO explores protoplanetary disk evolution by studying gas and dust across various ages. This work focuses on ten evolved disks in Upper Scorpius, observed in dust continuum emission, CO and its isotopologues, and N$_2$H$^+$ with ALMA Bands 6 and 7. Disk radii, from the radial location enclosing 68% of the flux, are comparable to those in the younger Lupus region for both gas and dust tracers. However, solid masses are about an order of magnitude below those in Lupus and Ophiuchus, while the dust spectral index suggests some level of dust evolution. These empirical findings align with a combination of radial drift, dust trapping, and grain growth into larger bodies. A moderate correlation between CO and continuum fluxes suggests a link between gas and dust content, through the increased scatter compared to younger regions, possibly due to age variations, gas-to-dust ratio differences, or CO depletion. Additionally, the correlation between C$^{18}$O and N$_2$H$^+$ fluxes observed in Lupus persists in Upper Sco, indicating a relatively stable CO gas abundance over the Class II stage of disk evolution. In conclusion, the AGE-PRO survey of Upper Scorpius disks reveals intriguing trends in disk evolution. The findings point towards potential gas evolution and the presence of dust traps in these older disks. Future high-resolution observations are needed to confirm these possibilities and further refine our understanding of disk evolution and planet formation in older environments.

astro-ph.EP↗

arXiv2025

The ALMA Survey of Gas Evolution of PROtoplanetary Disks (AGE-PRO): XII. Extreme millimetre variability detected in a Class II disc

Variability of millimetre wavelength continuum emission from Class II protoplanetary disks is extremely rare, and when detected it is usually interpreted as originating from non-thermal emission mechanisms that relate to the host star itself rather than its disk. During observations made as part of the AGE-PRO ALMA Large program, significant variability in the brightness of the 2MASS J16202863-2442087 system was detected between individual executions. We report the observed properties of the variability detected at millimetre wavelengths and investigate potential driving mechanisms. To investigate the nature of the variability we construct a light curve from the continuum observations and analyse imaged constructed from both flaring and quiescent emission. We characterise the dust disk around the star through analysis in the image and visibility plane, and carry out kinematic analysis of the CO(2-1) emission from the gas disk. The continuum flux decays by a factor of 8 in less than an hour, and by a factor of 13 within 8 days. The peak brightness coincides with an expected brightness maximum extrapolated from the periodicity of previously observed optical variability. The flare is most likely the product of synchrotron emission in the close vicinity of the star. The nature of the millimetre flare closely resembles those detected in very close binary systems, and may be due to the interaction of magnetic fields in an as yet undetected binary. Alternatively if the central host is a single-star object, the flare may be due to the interaction of magnetic field loops at the stellar surface or a strong accretion burst.

astro-ph.EP↗