Portail HAL Nantes Université
Not a member yet
120169 research outputs found
Sort by
P1056 Effectiveness and Safety of Upadacitinib Induction Therapy in Real-World Setting for 234 Patients with Crohn’s Disease: A GETAID Multicentre Cohort Study
No full textInternational audienceBackgroundWhile upadacitinib has demonstrated its efficacy as an induction treatment for Crohn’s disease (CD) in two phase 3 randomized, placebo-controlled trials1, real-world data remain limited.MethodsFrom September 2022 to September 2024, all consecutive patients with refractory CD treated with once-daily upadacitinib 45 mg in 30 French and Belgian GETAID centres were retrospectively included. The primary endpoint was steroid-free clinical remission (SFCR) at week 12, defined as a Harvey-Bradshaw Index (HBI) score of < 4. Secondary endpoints included clinical response (decrease of ≥ 3 points in the HBI score and/or an HBI score < 4), clinical remission, biological remission (calprotectin ≤ 250 µg/g, or CRP ≤ 5 if calprotectin was unavailable), endoscopic or radiological (IUS and/or MRI) response, and adverse events (AEs).Results234 patients were included, all of whom had been previously exposed to at least one biologic (median 4, IQR[3-4]), and 125 (53.9%) had undergone prior intestinal resection (Table 1). At week 12, SFCR was observed in 107 patients (n=107/197, 54%), clinical response in 120 (n=120/190, 63%) and clinical remission in 111 (n=111/197, 56%). Ninety-two (n=92/179, 51%) achieved biological remission and endoscopic or radiological response was observed in 19 (n=19/40, 48%) patients (Figure 1). Respectively 28 (n=28/37, 76%) and 20 (n=20/37, 54%) of patients with articular EIMs achieved clinical response and remission. For cutaneous EIM, clinical response was achieved in 10 patients (n=10/11, 91%), and clinical remission was achieved in nine patients (n=9/11, 82%). In multivariate analysis, body mass index < 18.5 kg/m2 (OR=0.09, 95%CI: 0.01-0.33, p=0.002) and HBI > 7 (OR=0.24, 95%CI: 0.12-0.47, p<0.0001) were associated with lower rates of SFCR at W12 while the number of prior biologics did not influence SFCR (OR = 0.72, 95% CI: 0.35-1.48; p = 0.36).At week 12, 35 (15%) patients discontinued upadacitinib (lack of effectiveness, n=31; AEs, n=2 and other, n=2). CD-related hospitalization was needed in 18 (7.7%) patients, and 4 (1.7%) underwent intestinal resection. Sixty-eight AEs occurred in 61 patients (26%), including 19 serious AEs, corresponding to 18 cases of CD exacerbation and one case of colonic EBV-associated lymphoproliferative disorder. Acne was observed in 25 (10.7%), justifying treatment discontinuation in one (0.4%).ConclusionIn this real-world cohort of highly refractory CD patients, upadacitinib induction resulted in a clinical response in about two-thirds of patients and SFCR in half of patients
Droit au renouvellement et validité du congé délivré par un mandataire social avant la publication de sa nomination
No full textInternational audienc
Modelling the thermal regime of perched talus slopes: The Gislá landslide (06/10/2020, Iceland)
No full textInternational audienc
Cover Feature: The ARTISTIC Battery Manufacturing Digitalization Initiative: From Fundamental Research to Industrialization (Batteries & Supercaps 1/2025)
No full textInternational audienceThe Cover Feature represents the whole ARTISTIC project workflow to optimize battery manufacturing process parameters. Synthetic data (produced by the physics-based manufacturing modeling chain) and experimental data are used to train surrogate models by using different machine learning techniques at the different manufacturing stages: mixing & slurry, coating & drying, calendering, electrolyte filling and performance. Then, optimizers, such as Bayesian, are used to determine the best input parameters to optimize output battery properties. More information can be found in the Concept by A. A. Franco and co-workers (DOI: 10.1002/batt.202400385)
Backtracking Enabled Transformers
No full textInternational audienceWhile Transformer models have revolutionized natural language processing, their reliance on statistical patterns and attention mechanisms limits their inherent logical reasoning capabilities. This leads to inconsistent answers when presented with varied problems, as predictions are based solely on correlation analysis within text sequences. Generative AI, powered by these models, struggles with self-correction and logical consistency, especially in complex reasoning tasks. Techniques like Chain of Thought alleviate these issues but become ineffective once an error is introduced in the reasoning process.To overcome these limitations, we propose a groundbreaking approach: fine-tuning backtracking-enabled Transformers on existing open-source large language models (LLMs). By equipping LLMs with symbolic reasoning capabilities without extensive retraining, we enhance the logical inference abilities of language models while enabling dynamic error correction during the reasoning process. This innovation has profound implications for AI applications that require complex logical reasoning, such as scientific research, legal analysis, and knowledge engineering. By incorporating backtracking, our modified Transformers can revisit and revise incorrect assumptions, ensuring more accurate and robust decision-making. This development paves the way for a new generation of LLMs that can navigate the complexities of logical reasoning with increased reliability, accuracy, and transparency.As a proof-of-concept, we present a Transformer designed to traverse tree structures effectively, showcasing the feasibility of backtracking in generative AI models. This foundational research aims to lay the groundwork for the creation of more intelligent, reliable, and trustworthy AI systems. The potential implications of this research are vast, particularly in fields that require precise, logical decision-making. By incorporating backtracking, we can create a new generation of language models that not only process vast amounts of data but also critically think through problems, correcting their own errors along the way. By enhancing the logical reasoning capabilities of these models, we can unlock new possibilities for human-AI collaboration and decision-making, driving innovation and progress in many fields.</div
AnCoGen: Analysis, Control and Generation of Speech with a Masked Autoencoder
No full textInternational audienceThis article introduces AnCoGen, a novel method that leverages a masked autoencoder to unify the analysis, control, and generation of speech signals within a single model. AnCoGen can analyze speech by estimating key attributes, such as speaker identity, pitch, content, loudness, signal-to-noise ratio, and clarity index. In addition, it can generate speech from these attributes and allow precise control of the synthesized speech by modifying them. Extensive experiments demonstrated the effectiveness of AnCoGen across speech analysisresynthesis, pitch estimation, pitch modification, and speech enhancement. Code and audio examples are available online
Studying charm hadronisation into baryons with azimuthal correlations of with charged particles in pp collisions at TeV
No full textInternational audienceThe distribution of angular correlations between prompt charm hadrons and primary charged particles in pp collisions is sensitive to the charm-quark hadronisation process. In this letter, charm-baryon correlations are measured for the first time by studying the azimuthal-angle difference between charged particles and prompt baryons produced in pp collisions at a centre-of-mass energy TeV, with the ALICE detector. baryons are reconstructed at midrapidity ( GeV/ and pseudorapidity . For GeV/, the comparison with published measurements of D-meson and charged-particle correlations in the same collision system hints at a larger number of low-momentum particles associated with -baryon triggers than with D-meson triggers, both in the collinear and opposite directions with respect to the trigger particle. These differences can be quantified by the comparison of the properties of the near- and away-side correlation peaks, and are not reproduced by predictions of various Monte Carlo event generators, generally underpredicting the associated particle yields at GeV/. This tension between -baryon and D-meson associated peak yields could suggest a modified fragmentation of the charm quark, or a different hadronisation process, when a charm baryon is produced in the final state
Neutrinoless Double Beta Decay Sensitivity of the XLZD Rare Event Observatory
No full textInternational audienceThe XLZD collaboration is developing a two-phase xenon time projection chamber with an active mass of 60 to 80 t capable of probing the remaining WIMP-nucleon interaction parameter space down to the so-called neutrino fog. In this work we show that, based on the performance of currently operating detectors using the same technology and a realistic reduction of radioactivity in detector materials, such an experiment will also be able to competitively search for neutrinoless double beta decay in Xe using a natural-abundance xenon target. XLZD can reach a 3 discovery potential half-life of 5.710 yr (and a 90% CL exclusion of 1.310 yr) with 10 years of data taking, corresponding to a Majorana mass range of 7.3-31.3 meV (4.8-20.5 meV). XLZD will thus exclude the inverted neutrino mass ordering parameter space and will start to probe the normal ordering region for most of the nuclear matrix elements commonly considered by the community