HAL Portal UDL Université de Lyon
Not a member yet
327890 research outputs found
Sort by
Embodied speech: Sensorimotor contributions to native and non-native phoneme processing and learning
No full textInternational audienceLearning to recognize and produce foreign speech sounds can be challenging, particularly when only subtle differences distinguish these new sounds from phonemes in the native language. Functional neuroimaging evidence shows that the motor cortex is involved in speech production and in perceptual phonemic processing. This highlights the embodied nature of speech perception, predicting the potential benefits of sensorimotor-based training approaches to enhance the acquisition of foreign speech sounds. Hence, here we first review current findings on the motor contribution to not only native but also non-native phoneme perception. Available evidence has established that motor cortical activity especially shows up under non-optimal perceptual conditions, such as when native phonemes are degraded by noise or when listeners perceive non-native speech sounds. Drawing upon this evidence, we then review training paradigms that have been developed for learning foreign phonemes, with a special emphasis on those embedding manual gestures as cues to represent phonetic features of the to-be-learned speech sounds. By pointing to both strengths and caveats of available studies, this review allows to delineate a clear framework and opens perspectives to optimize foreign phoneme learning, and ultimately support perception and production.</div
The Model's Language Matters: A Comparative Privacy Analysis of LLMs
No full textInternational audienceLarge Language Models (LLMs) are increasingly deployed in multilingual settings that process sensitive data, yet their scale and linguistic variability can amplify privacy risks. While prior privacy evaluations focus predominantly on English, we investigate how language structure shapes privacy leakage in LLMs trained on English, Spanish, French, and Italian medical corpora. We quantify six corpus-level linguistic indicators and evaluate vulnerability under three attack families: extraction, counterfactual memorization, and membership inference. Across languages, we find that leakage systematically tracks structural properties: Italian exhibits the strongest exposure, consistent with its highest redundancy and longer lexical units, whereas English shows the clearest membership separability, aligning with its higher syntactic entropy and stronger surface-identifiable cues. In contrast, French and Spanish remain comparatively more resilient overall, aided by higher morphological complexity. These results provide quantitative evidence that language matters for privacy leakage, motivating language-aware and structure-adaptive privacy-preserving mechanisms for multilingual LLM deployments
Redécouvrir l’ouvrage <i>Un travail à soi</i>
No full textInternational audienceLe premier séminaire transversal du Centre Max Weber de l’année 2026 se tiendra le 09 janvier. Cette séance sera dédiée à la réédition du livre de Phillipe Bernoux Un travail à soi. Nous avons souhaité faire de ce séminaire un lieu de discussion autour des travaux de Philippe Bernoux et d’échanges avec celles et ceux qui n’ont pu participer à la réédition de l’ouvrage, mais qui pourront témoigner du travail engagé avec lui. - Michèle Dupré et Michel Lallemen
Quantitative sensory testing of pain in osteoporosis: a pilot randomized clinical trial with magnesium supplementation
No full textInternational audienc
Déploiement et Maillage de stations instrumentées en IoT sur le territoire pour diverses études_AG AnaEE 2026 "ECOLOGGING"
No full textInternational audienceIn this poster, we present the deployments of the different types of instrumented stations developed within the framework of the "ECOLOGGING” project, carried out during the year 2024/2025, as well as the main developments, the results obtained, and the conclusions. The “ECOLOGGING” stations are deployed according to the nature of the studies conducted and the associated specific scientific needs.Dans ce poster, nous présentons les déploiements des différents types de stations instrumentées développées dans le cadre du projet "ECOLOGGING", réalisés au cours de l’année 2024/2025, ainsi que les principales évolutions, les résultats obtenus et les conclusions. Les stations ECOLOGGING sont déployées en fonction de la nature des études menées et des besoins scientifiques spécifiques associés
Les facteurs de choix du référentiel de consolidation des groupes français cotés sur Euronext Growth
No full textInternational audiencePourquoi conserver un règlement français de consolidation autonome par rapport aux normes IFRS ? En s’appuyant sur la théorie néo-institutionnelle et la théorie de la tétranormalisation, cet article met en évidence les contraintes techniques et sociales en matière de choix du référentiel de consolidation des groupes françaiscotés sur le marché non réglementé Euronext Growth. Le règlement ANC 2020-01 apparaît comme une solution économique, adaptée aux spécificités des PME tandis que les normes IFRS seraient utilisées comme un outil de légitimation vis-à-vis des parties prenantes
Reconstruction of atmospheric neutrinos in DUNE's horizontal-drift far-detector module
No full textInternational audienceThis paper reports on the capabilities in reconstructing and identifying atmospheric neutrino interactions in one of the Deep Underground Neutrino Experiment's (DUNE) far detector modules, a liquid argon time projection chamber (LArTPC) with horizontal drift (FD-HD) of ionization electrons. The reconstruction is based upon the workflow developed for DUNE's long-baseline oscillation analysis, with some necessary machine-learning models' retraining and the addition of features relevant only to atmospheric neutrinos such as the neutrino direction reconstruction. Where relevant, the impact of the detection of the charged particles of the hadronic system is emphasized, and comparisons are carried out between the case when lepton-only information is considered in the reconstruction (as is the case for many neutrino oscillation experiments), versus when all particles identified in the LArTPC were included. Three neutrino direction reconstruction methods have been developed and studied for the atmospheric analyses: using lepton-only information, using all reconstructed particles, and using only correlations from reconstructed hits. The results indicate that incorporating more than just lepton information significantly improves the resolution of both neutrino direction and energy reconstruction. The angle reconstruction algorithms developed in this work result in no strong dependence on particle direction for reconstruction efficiencies or neutrino flavor identification. This comprehensive review of the reconstruction of atmospheric neutrinos in DUNE's FD-HD LArTPC is the first step towards developing a first neutrino oscillation sensitivity analysis, which will ready DUNE for its first measurements
The Lazuli Space Observatory: Architecture & Capabilities
No full textInternational audienceThe Lazuli Space Observatory is a 3-meter aperture astronomical facility designed for rapid-response observations and precision astrophysics across visible to near-infrared wavelengths (400-1700 nm bandpass). An off-axis, freeform telescope delivers diffraction-limited image quality (Strehl 0.8 at 633 nm) to three instruments across a wide, flat focal plane. The three instruments provide complementary capabilities: a Wide-field Context Camera (WCC) delivers multi-band imaging over a 35' 12' footprint with high-cadence photometry; an Integral Field Spectrograph (IFS) provides continuous 400-1700 nm spectroscopy at R 100-500 for stable spectrophotometry; and an ExtraSolar Coronagraph (ESC) enables high-contrast imaging expected to reach raw contrasts of and post-processed contrasts approaching . Operating from a 3:1 lunar-resonant orbit, Lazuli will respond to targets of opportunity in under four hours--a programmatic requirement designed to enable routine temporal responsiveness that is unprecedented for a space telescope of this size. Lazuli's technical capabilities are shaped around three broad science areas: (1) time-domain and multi-messenger astronomy, (2) stars and planets, and (3) cosmology. These capabilities enable a potent mix of science spanning gravitational wave counterpart characterization, fast-evolving transients, Type Ia supernova cosmology, high-contrast exoplanet imaging, and spectroscopy of exoplanet atmospheres. While these areas guide the observatory design, Lazuli is conceived as a general-purpose facility capable of supporting a wide range of astrophysical investigations, with open time for the global community. We describe the observatory architecture and capabilities in the preliminary design phase, with science operations anticipated following a rapid development cycle from concept to launch
Euclid: Improving redshift distribution reconstruction using a deep-to-wide transfer function
No full textInternational audienceThe Euclid mission seeks to understand the Universe expansion history and the nature of dark energy, which requires a very accurate estimate of redshift distribution. Achieving this accuracy relies on reference samples with spectroscopic redshifts, together with a procedure to match them to survey sources for which only photometric redshifts are available. One important source of systematic uncertainty is the mismatch in photometric properties between galaxies in the Euclid survey and the reference objects. We develop a method to degrade the photometry of objects with deep photometry to match the properties of any shallower survey in the multi-band photometric space, preserving all the correlations between the fluxes and their uncertainties. We compare our transfer method with more demanding image-based methods, such as Balrog from the Dark Energy Survey Collaboration. According to metrics, our method outperforms Balrog. We implement it in the redshift distribution reconstruction, based on the self-organising map approach of arXiv:1509.03318, and test it using a realistic sample from the Euclid Flagship Simulation. We find that the key ingredient is to ensure that the reference objects are distributed in the colour space the same way as the wide-survey objects, which can be efficiently achieved with our transfer method. In our best implementation, the mean redshift biases are consistently reduced across the tomographic bins, bringing a significant fraction of them within the Euclid accuracy requirements in all tomographic bins. Equally importantly, the tests allow us to pinpoint which step in the calibration pipeline has the strongest impact on achieving the required accuracy. Our approach also reproduces the overall redshift distributions, which are crucial for applications such as angular clustering
