Source author record

Y. charles

Y. charles appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

astro-ph.IM astro-ph.CO astro-ph.EP astro-ph.GA Computation and Language Computer Vision physics.flu-dyn

Catalog footprint

What is connected

6works

7topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

BabyVision: Visual Reasoning Beyond Language

While humans develop core visual skills long before acquiring language, contemporary Multimodal LLMs (MLLMs) still rely heavily on linguistic priors to compensate for their fragile visual understanding. We uncovered a crucial fact: state-of-the-art MLLMs consistently fail on basic visual tasks that humans, even 3-year-olds, can solve effortlessly. To systematically investigate this gap, we introduce BabyVision, a benchmark designed to assess core visual abilities independent of linguistic knowledge for MLLMs. BabyVision spans a wide range of tasks, with 388 items divided into 22 subclasses across four key categories. Empirical results and human evaluation reveal that leading MLLMs perform significantly below human baselines. Gemini3-Pro-Preview scores 49.7, lagging behind 6-year-old humans and falling well behind the average adult score of 94.1. These results show despite excelling in knowledge-heavy evaluations, current MLLMs still lack fundamental visual primitives. Progress in BabyVision represents a step toward human-level visual perception and reasoning capabilities. We also explore solving visual reasoning with generation models by proposing BabyVision-Gen and automatic evaluation toolkit. Our code and benchmark data are released at https://github.com/UniPat-AI/BabyVision for reproduction.

preprint2026arXiv

Euclid: Galaxy SED reconstruction in the PHZ processing function: impact on the PSF and the role of medium-band filters

Weak lensing surveys require accurate correction for the point spread function (PSF) when measuring galaxy shapes. For a diffraction-limited PSF, as arises in space-based missions, this correction depends on each galaxy SED. In the Euclid mission, galaxy SED reconstruction, a tasks of the photometric-redshift processing function (PHZ PF), relies on broad- and medium-band ancillary photometry. The limited wavelength sampling of the Euclid VIS passband and signal-to-noise ratio may affect the reconstruction accuracy and translate into biases in the weak lensing measurements. In this study, we present the methodology, which is employed in the Euclid PHZ PF, for reconstructing galaxy SEDs at 55 wavelengths, sampling the VIS passband every 10 nm, and we assess whether it fulfils the accuracy requirements imposed on the Euclid PSF model. We employ both physics- and data-driven methods, focusing on a new approach of template-based flux correction and Gaussian processes, and we introduce an SED metric whose bias propagates into PSF quadrupole moment errors. Our findings demonstrate that Gaussian processes and template fitting meet the requirements only in specific, but complementary, redshift intervals. We therefore propose a hybrid approach, which leverages both methods. This solution proves to be effective in meeting the Euclid accuracy requirements for most of the redshift range of the survey. Finally, we investigate the impact on the SED reconstruction of a new set of 16 evenly-spaced medium-band filters for the Subaru telescope, providing quasi-spectroscopic coverage of the VIS passband. This study shows promising results, ensuring accurate SED reconstruction and meeting the mission PSF requirements. This work thus provides not only the methodological foundation of galaxy SED reconstruction in the Euclid PHZ PF, but also a roadmap for future improvements using a new medium-band survey.

preprint2022arXiv

Connecting SPHERE and CRIRES+ for the characterisation of young exoplanets at high spectral resolution: status update of VLT/HiRISE

New generation exoplanet imagers on large ground-based telescopes are highly optimised for the detection of young giant exoplanets in the near-infrared, but they are intrinsically limited for their characterisation by the low spectral resolution of their integral field spectrographs ($R<100$). High-dispersion spectroscopy at $R \gg 10^4$ would be a powerful tool for the characterisation of these planets, but there is currently no high-resolution spectrograph with extreme adaptive optics and coronagraphy that would enable such characterisation. With project HiRISE we propose to use fiber coupling to combine the capabilities of two flagship instruments at the Very Large Telescope in Chile: the exoplanet imager SPHERE and the high-resolution spectrograph CRIRES+. The coupling will be implemented at the telescope in early 2023. We provide a general overview of the implementation of HiRISE, of its assembly, integration and testing (AIT) phase in Europe, and a brief assessment of its expected performance based on the final hardware.

preprint2020arXiv

Direct characterization of young giant exoplanets at high spectral resolution by coupling SPHERE and CRIRES+

Studies of atmospheres of directly imaged exoplanets with high-resolution spectrographs have shown that their characterization is predominantly limited by noise on the stellar halo at the location of the studied exoplanet. An instrumental combination of high-contrast imaging and high spectral resolution that suppresses this noise and resolves the spectral lines can therefore yield higher quality spectra. We study the performance of the proposed HiRISE fiber coupling between the SPHERE and CRIRES+ at the VLT for spectral characterization of directly imaged planets. Using end-to-end simulations of HiRISE we determine the S/N of the detection of molecular species for known exoplanets in $H$ and $K$ bands, and compare them to CRIRES+. We investigate the ultimate detection limits of HiRISE as a function of stellar magnitude, and we quantify the impact of different coronagraphs and of the system transmission. We find that HiRISE largely outperforms CRIRES+ for companions around bright hosts like $β$ Pic or 51 Eri. For an $H=3.5$ host, we observe a gain of a factor of up to 16 in observing time with HiRISE to reach the same S/N on a companion at 200 mas. More generally, HiRISE provides better performance than CRIRES+ in two-hour integration times between 50-350 mas for hosts with $H<8.5$ and between 50-700 mas for $H<7$. For fainter hosts like PDS 70 and HIP 65426, no significant improvements are observed. We find that using no coronagraph yields the best S/N when characterizing known exoplanets due to higher transmission and fiber-based starlight suppression. We demonstrate that the overall transmission of the system is in fact the main driver of performance. Finally, we show that HiRISE outperforms the best detection limits of SPHERE for bright stars, opening major possibilities for the characterization of future planetary companions detected by other techniques.

preprint2019arXiv

Euclid preparation: VI. Verifying the Performance of Cosmic Shear Experiments

Our aim is to quantify the impact of systematic effects on the inference of cosmological parameters from cosmic shear. We present an end-to-end approach that introduces sources of bias in a modelled weak lensing survey on a galaxy-by-galaxy level. Residual biases are propagated through a pipeline from galaxy properties (one end) through to cosmic shear power spectra and cosmological parameter estimates (the other end), to quantify how imperfect knowledge of the pipeline changes the maximum likelihood values of dark energy parameters. We quantify the impact of an imperfect correction for charge transfer inefficiency (CTI) and modelling uncertainties of the point spread function (PSF) for Euclid, and find that the biases introduced can be corrected to acceptable levels.

preprint2013arXiv

Direct and inverse pumping in flows with homogeneous and non-homogeneous swirl

The conditions in which meridional recirculations appear in swirling flows above a fixed wall are analysed. In the classical Bodewädt problem, where the swirl tends towards an aysmptotic value away from the wall, the well-known "tea-cup effect" drives a flow away from the plate at the centre of the vortex. Simple dimensional arguments applied to a single vortex show that if the intensity of the swirl decreases away from the wall, the sense of the recirculation can be inverted, and that the associated flow rate scales with the swirl gradient. Only if the flow is quasi-2D, does the classical tea-cup effect take place. This basic theory is confirmed by numerical simulations of a square array of steady, electrically driven vortices. Experiments in the turbulent regimes of the same configuration reveal that these mechanisms are active in the average flow and in its fluctuating part. The mechanisms singled out in this letter provide an explanation for previously observed phenomena in electrolyte flows. They also put forward a possible mechanism for the generation of helicity in flows close to two-dimensionality, which plays a key role in the transition between 2D and 3D turbulence.