Abstract
Context. The center of our Galaxy hosts Sagittarius A*, which is a supermassive compact object of similar to 4.3 & times; 10(6) solar masses and is usually associated with a black hole. Nevertheless, black holes possess a central singularity that is considered unphysical, and an event horizon that leads to loss of unitarity in a quantum description of the system. To address these theoretical inconsistencies, alternative models, collectively known as exotic compact objects, have been proposed. Aims. We investigate the potential detectability of signatures associated with nonrotating exotic compact objects (ECOs) within the dataset of Sgr A* polarized flares as observed through GRAVITY and the upcoming GRAVITY+. Methods. We examined a total of eight distinct metrics that originate from four different categories of static and spherically symmetric compact objects: black holes, boson stars, fluid spheres, and gravastars. Our approach involved using a toy model that orbits the compact object in the equatorial plane at the Schwarzschild-Keplerian velocity. Using simulated astrometric and polarimetric data with current GRAVITY uncertainties as well as improved flux uncertainties expected for the GRAVITY+ upgrade, we fit the datasets across all metrics we examined. We evaluated the detectability of the metric for each dataset based on the resulting chi(2)(red) and Bayesian information criteria-based Bayes factors. Results. Plunge-through images of ECOs affect polarization and astrometry in a distinguishable way from the spin of a Kerr black hole. With GRAVITY's current uncertainties, none of the metrics models are discernible. However, when the data are modeled within a compact boson star background, the corresponding best fit is sufficiently superior to the Kerr fit to rule out the latter. We examined the best expected enhanced flux uncertainties and discovered that a fourfold increase in flux sensitivity enables the detection of some of the exotic compact object models we investigated. The signals of the others are too close to each other to be distinguishable. However, with the GRAVITY+ flux uncertainties, when the data are produced using an ECO model, the best-fit ECO model is significantly preferred (with a BIC-based Bayes factor exceeding two) over the best fit in the Kerr metric, such that the latter can be ruled out. Nevertheless, enhancing the astrophysical complexity of the hot-spot model might diminishes these outcomes. Conclusions. With the improved sensitivity of GRAVITY+, we expect to be able to determine whether Sgr A* is a Kerr black hole or some form of exotic compact object, although we will not be able to identify the specific ECO models that describe Sgr A* best.

Abstract
WISPIT 2 is a nearby young star with a multiringed disk that was recently confirmed to host a similar to 4.9 MJup gas giant planet embedded in a large (60 au) gap at a radial separation of 57 au from the host star. We confirm and characterize a second, close-in planet in the WISPIT 2 system using a combination of new Very Large Telescope/SPHERE H-band dual-polarization imaging and VLTI/GRAVITY K-band interferometric observations of the WISPIT 2 system. The GRAVITY detection is consistent with a point-like source while its extracted K-band spectrum shows CO band-head absorption at 2.3 mu m and a continuum shape consistent with a young giant planet. From the GRAVITY data, we extract a medium resolution K-band spectrum of the companion and fit atmospheric model grids using the species tool with nested sampling to constrain its effective temperature, radius, and luminosity. We infer Teff of 1500-2600 K, a radius of 0.91-2.2 RJup, and a luminosity of (-3.47)-(-3.63). Comparison with evolutionary tracks implies a mass range of 8-12 MJup, approximately twice as massive as the previously confirmed WISPIT 2b. The astrometry rules out a background source and marginally detects orbital motion of WISPIT 2 c, which needs further follow-up observations for confirmation. WISPIT 2 now becomes an analog to PDS 70, offering a second laboratory for studying the formation and early evolution of a multiplanet system within its natal disk.

Abstract
The centre of the Galaxy harbours a supermassive black hole, Sgr A*, which is surrounded by a massive star cluster known as the S-cluster. The most extensively studied star in this cluster is the B-type main-sequence S2 star (also known as S0-2). These types of stars are commonly found in binary systems in the Galactic field, but observations do not seem to detect a companion to S2. This absence may be attributed to observational biases or to a dynamically hostile environment caused by phenomena such as tidal disruption or mergers. Using a N-body code with first-order post-Newtonian corrections, we investigate whether S2 can host a stellar or planetary companion. We perform 105 simulations adopting uniform distributions for the orbital elements of the companion. Our results show that companions may exist for orbital periods shorter than 100 days, eccentricities below 0.8, and across the full range of mutual inclinations. The number of surviving companions increases with shorter orbital periods, lower eccentricities, and nearly coplanar orbits. We also find that the disruption mechanisms include mergers driven by Lidov-Kozai cycles and breakups that occur when the companion surpasses the Hill radius of its orbit. Finally, we find that the presence of a companion would alter S2's astrometric signal by no more than 5 mu as. Current radial-velocity detection limits constrain viable stellar binary configurations to approximately 4.4% of the simulated cases. Including astrometric limits reduces to 4.3%. Imposing an additional constraint that any companion must have a mass less than or similar to 2 M-circle dot (otherwise it would be visible) narrows the fraction of undetectable stellar binaries to just 3.0%.

Abstract
Optical coherence tomography angiography (OCTA), known for its high-resolution and noninvasive imaging capability, has become a key modality for visualizing retinal vasculature. Accurate and automated segmentation of capillaries, arteries, veins, and foveal avascular zone in OCTA images is essential for quantitative analysis and disease assessment. In this paper, we propose a depth enhanced cascaded framework specifically designed for multi-class OCTA segmentation. Our method investigates the spatial distribution of vasculature in retinal images and integrates a novel self-supervised depth prediction module to learn implicit depth cues from volumetric data, thereby improving the discrimination of overlapping vascular layers. In addition, we design two topology-aware loss functions that explicitly encourage structural integrity and continuity of vessel segmentation, particularly at bifurcations and endpoints. Experiments on the OCTA-6 mm and OCTA-3 mm datasets demonstrate that our method outperforms existing state-of-the-art approaches, with mIoU gains of around 2% over prior method, IPNv2, thereby highlighting enhanced segmentation accuracy and vascular topology preservation.

Abstract
Highlights What are the main findings? PDAM improves hyperspectral image classification robustness under noisy labels. Prototype-guided dynamic masking and reconstruction outperform competing methods across multiple noise levels and datasets. What are the implications of the main findings? Prototype-based reliability estimation is effective for identifying trustworthy supervision in noisy hyperspectral data. Reliability-aware masking offers a practical strategy for robust remote-sensing learning with corrupted annotations.Highlights What are the main findings? PDAM improves hyperspectral image classification robustness under noisy labels. Prototype-guided dynamic masking and reconstruction outperform competing methods across multiple noise levels and datasets. What are the implications of the main findings? Prototype-based reliability estimation is effective for identifying trustworthy supervision in noisy hyperspectral data. Reliability-aware masking offers a practical strategy for robust remote-sensing learning with corrupted annotations.Abstract Existing noisy-label hyperspectral image classification (HSIC) methods usually address clean sample selection and representation regularization as separate problems, although the reliability of observed labels varies substantially across samples in hyperspectral data. This issue is amplified by mixed pixels, boundary ambiguity, spectral overlap, and limited labeled samples, which make hard clean samples difficult to distinguish from mislabeled ones. We therefore propose PDAM, a sample-reliability-guided training framework for noisy-label HSIC. The method first estimates feature-space class consistency by comparing each sample with the prototype of its observed class and converting this consistency into a reliability probability with a Gaussian mixture model. To reduce conservative false negatives, matched high-confidence selection is further used to recover hard but correctly labeled samples. The resulting reliability estimate then determines how strongly the observed label is trusted through target refinement and how strongly the input is perturbed through reliability-guided masking. Finally, masked reconstruction provides label-independent structural regularization so that uncertain samples can still contribute to spectral-spatial representation learning. Under the evaluated synthetic symmetric noise settings on the University of Pavia (UP), Salinas Valley (SV), and Kennedy Space Center (KSC) datasets, PDAM achieves the best OA and Kappa in most reported comparisons and improves robustness under both moderate and severe noise. At 30% noise, PDAM reaches 97.30% OA on UP, 98.13% OA on SV, and 95.37% OA on KSC. Ablation studies further support the necessity of reliability estimation, hard clean sample recovery, and reliability-guided supervision and regularization within this unified training mechanism.

Abstract