Search

We report the pinching of an elongated liquid droplet that is confined in a rectangular microchannel. The droplet pinching is induced by a local variation of the channel topography and can lead to its break-up. The modification of the channel topography is either induced by a reversible local dilation of the channel bottom wall or by a confinement gradient that is irreversibly modelled in the channel. Interestingly, in both cases, a few micrometres contraction of the channel height leads to the pinching of the droplet in the direction of the channel width. If this pinching is such that the droplet is no longer confined at the level of its deformation, {\it i.e.} the liquid thread becomes 3D, the droplet splits in two. By minimizing the surface energy of the droplet under the channel confinement gradient, we are able to predict its pinching or its break-up, if any. The dynamic of the droplet deformation is then captured by a semi-analytical model, assuming that most of the viscous dissipation is located in the gutters at the four corners of the channel. Conversely, we show that this kinetic model can be used to extract the average topographic variation of the channel, down to few micrometers, by measuring the droplet pinching dynamics under a classical optical microscope.

Materials imperfections in Nniobium based superconducting quantum circuits, in particular, two-level-system (TLS) defects, are a major source of decoherence, ultimately limiting the performance of quantum computation and sensing. Thus, identifying and understanding the microscopic origin of possible TLS defects in these devices and developing strategies to eliminate them is key to superconducting qubit performance improvement. In this paper, we demonstrate the reduction of two-level system losses in three-dimensional superconducting radio frequency (SRF) niobium resonators by a 10-hour high vacuum (HV) heat treatment at 650{\deg}C, even after exposure to air and high pressure rinsing (HPR). By probing the effect of this annealing on niobium samples using X-ray photoelectron spectroscopy (XPS) and high-resolution scanning transmission electron microscopy (STEM), we witness an alteration of the native oxide composition re-grown after air exposure and HPR and the creation of nano-scale crystalline oxide regions, which correlates with the measured tenfold quality factor enhancement at low fields of the 1.3 GHz niobium resonator.

IJCLab has been leading the development and deployment of low-$\beta$ Superconducting Radio Frequency (SRF) cavities for proton and heavy ion accelerators. We are launching an electron accelerator project for sustainable Energy Recovery Linac (iSAS/PERLE) with state-of-the-art SRF cavities at 800~MHz. Our proposal includes advanced heat treatment of such cavities to reach an excellent quality factor of $3\times 10^{10}$ at $22$~MV/m. In this paper, we overview the status of this activity., Comment: Contribution to LCWS2024 proceedings

Matching patients to clinical trials demands a systematic and reasoned interpretation of documents which require significant expert-level background knowledge, over a complex set of well-defined eligibility criteria. Moreover, this interpretation process needs to operate at scale, over vast knowledge bases of trials. In this paper, we propose a scalable method that extends the capabilities of LLMs in the direction of systematizing the reasoning over sets of medical eligibility criteria, evaluating it in the context of real-world cases. The proposed method overlays a Set-guided reasoning method for LLMs. The proposed framework is evaluated on TREC 2022 Clinical Trials, achieving results superior to the state-of-the-art: NDCG@10 of 0.693 and Precision@10 of 0.73.

Temperature gradients represent energy sources that can be harvested to generate steady reaction or transport fluxes. Technological developments could lead to the transfer of free energy from heat sources and sinks to chemical systems for the purpose of extraction, thermal batteries, or nonequilibrium synthesis. We present a theoretical study of 1D chemical systems subjected to temperature gradients, for sustaining nonequilibrium chemical fluxes. A complete theoretical framework describes the behavior of the system induced by various temperature profiles. An exact mathematical derivation was established for a simple two-compartment model and was generalized to arbitrary reaction-diffusion systems based on numerical models. An experimental system was eventually scaled and tuned to optimize either nonequilibrium chemical transport or reaction. The relevant parameters for this description were identified; they focused on the system symmetry for chemical reaction and transport. Nonequilibrium thermodynamic approaches lead to a description analogous to electric circuits. Temperature gradients lead to the onset of a steady chemical force, which maintains steady reaction-diffusion fluxes moderated by chemical resistance. The system activity was then assessed using the entropy production rate as a measure of its dissipated power. The chemical characteristics of the system can be tuned for general optimization of the nonequilibrium state or for the specific optimization of either transport or reaction processes. The shape of the temperature gradient can be tailored to precisely control the spatial localization of active processes, targeting either precise spatial localization or propagation over large areas. The resulting temperature-driven chemical system can in turn be used to drive secondary processes into either nonequilibrium reaction fluxes or concentration gradients., Comment: Maint text: 17 pages, 5 figures, 2 tables. Supplementary Information: 28 page, 5 figures. Submitted to the Physical Review E

Microfluidic trapping arrays have proven to be efficient tools for various applications that require working at the single-cell level, such as cell-cell communication or fusion. Although several hydrodynamic trapping devices have already been optimised, two-dimensional (2D) single-layer trapping arrays with high trap densities remain partially inefficient. Specifically, many traps remain empty, even after prolonged injection, which drastically reduces the number of samples available for post-treatment. These unfilled traps result from the symmetrical nature of the flow around the traps, and breaking this symmetry enhances capture efficiency. In this study, we use a numerical approachto show that optimal geometries can significantly increase filling efficiency and a preliminary experimental test confirming our approach is provided. We show that these improvements are achieved by promoting lateral dispersion of particles, facilitated either through an optimised oblique flow or by introducing disorder into the spatial arrangement of traps without specific inlet/outlet adjustment., Comment: 25 pages, 11 figures, 1 table. Supplementary information included in two sections instead of a supplementary file

This study investigates the use of Atomic Layer deposition (ALD) to mitigate multipacting phenomena inside superconducting radio frequency (SRF) cavities used in particle accelerators. The unique ALD capability to control the film thickness down to the atomic level on arbitrary complex shape objects enable the fine tuning of TiN film resistivity and total electron emission yield (TEEY) from coupons to devices. This level of control allows us to adequately choose a TiN film thickness that provide both a high resistivity to prevent Ohmic losses and low TEEY to mitigate multipacting for the application of interest. The methodology presented in this work can be scaled to other domain and devices subject to RF fields in vacuum and sensitive to multipacting or electron discharge processes with their own requirements in resistivities and TEEY values

Next basket recommendation (NBR) is a special type of sequential recommendation that is increasingly receiving attention. So far, most NBR studies have focused on optimizing the accuracy of the recommendation, whereas optimizing for beyond-accuracy metrics, e.g., item fairness and diversity remains largely unexplored. Recent studies into NBR have found a substantial performance difference between recommending repeat items and explore items. Repeat items contribute most of the users' perceived accuracy compared with explore items. Informed by these findings, we identify a potential "short-cut" to optimize for beyond-accuracy metrics while maintaining high accuracy. To leverage and verify the existence of such short-cuts, we propose a plug-and-play two-step repetition-exploration (TREx) framework that treats repeat items and explores items separately, where we design a simple yet highly effective repetition module to ensure high accuracy, while two exploration modules target optimizing only beyond-accuracy metrics. Experiments are performed on two widely-used datasets w.r.t. a range of beyond-accuracy metrics, viz. five fairness metrics and three diversity metrics. Our experimental results verify the effectiveness of TREx. Prima facie, this appears to be good news: we can achieve high accuracy and improved beyond-accuracy metrics at the same time. However, we argue that the real-world value of our algorithmic solution, TREx, is likely to be limited and reflect on the reasonableness of the evaluation setup. We end up challenging existing evaluation paradigms, particularly in the context of beyond-accuracy metrics, and provide insights for researchers to navigate potential pitfalls and determine reasonable metrics to consider when optimizing for accuracy and beyond-accuracy metrics., Comment: To appear at SIGIR'24

Large Language Models (LLMs) are at the forefront of NLP achievements but fall short in dealing with shortcut learning, factual inconsistency, and vulnerability to adversarial inputs.These shortcomings are especially critical in medical contexts, where they can misrepresent actual model capabilities. Addressing this, we present SemEval-2024 Task 2: Safe Biomedical Natural Language Inference for ClinicalTrials. Our contributions include the refined NLI4CT-P dataset (i.e., Natural Language Inference for Clinical Trials - Perturbed), designed to challenge LLMs with interventional and causal reasoning tasks, along with a comprehensive evaluation of methods and results for participant submissions. A total of 106 participants registered for the task contributing to over 1200 individual submissions and 25 system overview papers. This initiative aims to advance the robustness and applicability of NLI models in healthcare, ensuring safer and more dependable AI assistance in clinical decision-making. We anticipate that the dataset, models, and outcomes of this task can support future research in the field of biomedical NLI. The dataset, competition leaderboard, and website are publicly available.

Academic institutions and their staff use, adapt and create software. We're thinking of business tools used to carry out their mission: teaching management (Moodle) or subject teaching support (such as Maxima for formal calculus), for example. We're talking about software resulting from research work, designed by a researcher or a team as part of a research project (funded by ANR, Europe, etc. or not) or as a research service for a third party. These projects can last for decades (such as the Coq program proof assistant project, or the GPAC multimedia content distribution platform).We discuss why this software is produced, with what resources, the interest that institutions derive from it, what we call the ''valorization'' of software resulting from scientific research. The latter is multifaceted, as are the missions of scientific institutions: social value (contribution to the world heritage of knowledge), financial value (contracts), economic value (business creation), scientific value (publication), image value (visibility of the institution among target audiences: students, researchers, companies, prescribers)., Comment: in French language

The dataset focuses on Wikipedia users and contains information about demographic and socioeconomic characteristics of the respondents and their activity on Wikipedia. The data was collected using a questionnaire available online between June and July 2023. The link to the questionnaire was distributed via a banner published in 8 languages on the Wikipedia page. Filling out the questionnaire was voluntary and not incentivised in any way. The survey includes 200 questions about: what people were doing on Wikipedia before clicking the link to the questionnaire; how they use Wikipedia as readers (``professional'' and ``personal'' uses); their opinion on the quality, the thematic coverage, the importance of the encyclopaedia; the making of Wikipedia (how they think it is made, if they have ever contributed and how); their social, sport, artistic and cultural activities, both online and offline; their socio-economic characteristics including political beliefs, and trust propensities. More than 200 000 people opened the questionnaire, 100 332 started to answer, and constitute our dataset, and 10 576 finished it. Among other themes identified by future researchers, the dataset can be useful for advancing the research regarding the features of readers vs contributors of online commons, the relationship between trust, information, sources, and the use made of this information.

Distributional reinforcement learning (RL) has proven useful in multiple benchmarks as it enables approximating the full distribution of returns and makes a better use of environment samples. The commonly used quantile regression approach to distributional RL -- based on asymmetric $L_1$ losses -- provides a flexible and effective way of learning arbitrary return distributions. In practice, it is often improved by using a more efficient, hybrid asymmetric $L_1$-$L_2$ Huber loss for quantile regression. However, by doing so, distributional estimation guarantees vanish, and we empirically observe that the estimated distribution rapidly collapses to its mean. Indeed, asymmetric $L_2$ losses, corresponding to expectile regression, cannot be readily used for distributional temporal difference learning. Motivated by the efficiency of $L_2$-based learning, we propose to jointly learn expectiles and quantiles of the return distribution in a way that allows efficient learning while keeping an estimate of the full distribution of returns. We prove that our approach approximately learns the correct return distribution, and we benchmark a practical implementation on a toy example and at scale. On the Atari benchmark, our approach matches the performance of the Huber-based IQN-1 baseline after $200$M training frames but avoids distributional collapse and keeps estimates of the full distribution of returns.

How can we interpret and retrieve medical evidence to support clinical decisions? Clinical trial reports (CTR) amassed over the years contain indispensable information for the development of personalized medicine. However, it is practically infeasible to manually inspect over 400,000+ clinical trial reports in order to find the best evidence for experimental treatments. Natural Language Inference (NLI) offers a potential solution to this problem, by allowing the scalable computation of textual entailment. However, existing NLI models perform poorly on biomedical corpora, and previously published datasets fail to capture the full complexity of inference over CTRs. In this work, we present a novel resource to advance research on NLI for reasoning on CTRs. The resource includes two main tasks. Firstly, to determine the inference relation between a natural language statement, and a CTR. Secondly, to retrieve supporting facts to justify the predicted relation. We provide NLI4CT, a corpus of 2400 statements and CTRs, annotated for these tasks. Baselines on this corpus expose the limitations of existing NLI models, with 6 state-of-the-art NLI models achieving a maximum F1 score of 0.627. To the best of our knowledge, we are the first to design a task that covers the interpretation of full CTRs. To encourage further work on this challenging dataset, we make the corpus, competition leaderboard, website and code to replicate the baseline experiments available at: https://github.com/ai-systems/nli4ct, Comment: EMNLP 2023 Camera-ready, 15 pages

This paper describes the results of SemEval 2023 task 7 -- Multi-Evidence Natural Language Inference for Clinical Trial Data (NLI4CT) -- consisting of 2 tasks, a Natural Language Inference (NLI) task, and an evidence selection task on clinical trial data. The proposed challenges require multi-hop biomedical and numerical reasoning, which are of significant importance to the development of systems capable of large-scale interpretation and retrieval of medical evidence, to provide personalized evidence-based care. Task 1, the entailment task, received 643 submissions from 40 participants, and Task 2, the evidence selection task, received 364 submissions from 23 participants. The tasks are challenging, with the majority of submitted systems failing to significantly outperform the majority class baseline on the entailment task, and we observe significantly better performance on the evidence selection task than on the entailment task. Increasing the number of model parameters leads to a direct increase in performance, far more significant than the effect of biomedical pre-training. Future works could explore the limitations of large models for generalization and numerical inference, and investigate methods to augment clinical datasets to allow for more rigorous testing and to facilitate fine-tuning. We envisage that the dataset, models, and results of this task will be useful to the biomedical NLI and evidence retrieval communities. The dataset, competition leaderboard, and website are publicly available.