Search

Vision-language models (VLMs), which process image and text inputs, are increasingly integrated into chat assistants and other consumer AI applications. Without proper safeguards, however, VLMs may give harmful advice (e.g. how to self-harm) or encourage unsafe behaviours (e.g. to consume drugs). Despite these clear hazards, little work so far has evaluated VLM safety and the novel risks created by multimodal inputs. To address this gap, we introduce MSTS, a Multimodal Safety Test Suite for VLMs. MSTS comprises 400 test prompts across 40 fine-grained hazard categories. Each test prompt consists of a text and an image that only in combination reveal their full unsafe meaning. With MSTS, we find clear safety issues in several open VLMs. We also find some VLMs to be safe by accident, meaning that they are safe because they fail to understand even simple test prompts. We translate MSTS into ten languages, showing non-English prompts to increase the rate of unsafe model responses. We also show models to be safer when tested with text only rather than multimodal prompts. Finally, we explore the automation of VLM safety assessments, finding even the best safety classifiers to be lacking., Comment: under review

Superconducting-nanowire single-photon detectors (SNSPDs) have enabled the realization of several quantum optics technologies thanks to their high detection efficiency, low dark-counts, and fast recovery time. However, the widespread use of technologies such as linear optical quantum computing (LOQC), quasi-deterministic single photon sources and quantum repeaters requires faster detectors that can distinguish between different photon number states. Here, we report the fabrication of an SNSPD array composed of 14 independent pixels, achieving a system detection efficiency (SDE) of 90% in the telecom band. By reading each pixel of the array independently we show that the detector can detect telecom photons at 1.5 GHz with 45% absolute SDE. We exploit the dynamic PNR of the array to demonstrate accurate state reconstruction for different photon-number statistics for a wide range of light inputs, including operation with long-duration light pulses, as commonly obtained with some cavity-based sources. We show 2-photon and 3-photon fidelities of 74% and 57% respectively, which represent state-of-the-art results for fiber-coupled SNSPDs.

Quantum Key Distribution has made continuous progress over the last 20 years and is now commercially available. However, the secret key rates (SKR) are still limited to a few Mbps. Here, we present a custom multipixel superconducting nanowire single-photon detectors and fast acquisition and real-time key distillation electronics, removing two roadblocks and allowing an increase of the SKR of more than an order of magnitude. In combination with a simple 2.5 GHz clocked time-bin quantum key distribution system, we can generate secret keys at a rate of 64 Mbps over a distance of 10.0 km and at a rate of 3.0 Mbps over a distance of 102.4 km with real-time key distillation., Comment: 5 pages, 5 figures, submitted to Nature Photonics

Observation of interlayer, charge-transfer (CT) excitons in van der Waals heterostructures (vdWHs) based on 2D-2D systems has been well investigated. While conceptually interesting, these charge transfer excitons are highly delocalized and spatially localizing them requires twisting layers at very specific angles. This issue of localizing the CT excitons can be overcome via making mixed dimensional vdWHs (MDHs) where one of the components is a spatially quantum confined medium. Here, we demonstrate the formation of CT excitons in a 2D/quasi-2D system comprising MoSe2 and WSe2 monolayers and CdSe/CdS based core/shell nanoplates (NPLs). Spectral signatures of CT excitons in our MDHs were resolved locally at the 2D/single-NPL heterointerface using tip-enhanced photoluminescence (TEPL) at room temperature. By varying both the 2D material, the shell thickness of the NPLs, and applying out-of-plane electric field, the exciton resonance energy was tuned by up to 120 meV. Our finding is a significant step towards the realization of highly tunable MDH-based next generation photonic devices., Comment: 4 figures, 20 pages

Optical field localization at plasmonic tip-sample nanojunctions has enabled high spatial resolution chemical analysis through tip-enhanced linear optical spectroscopies, including Raman scattering and photoluminescence. Here, we illustrate that nonlinear optical processes, including parametric four-wave mixing (4WM), second harmonic/sum-frequency generation (SHG and SFG), and two-photon photoluminescence (TPPL), can be enhanced at plasmonic junctions and spatio-spectrally resolved simultaneously with few-nm spatial resolution under ambient conditions. More importantly, through a detailed analysis of our spectral nano-images, we find that the efficiencies of the local nonlinear signals are determined by sharp tip-sample junction resonances that vary over the few-nanometer length scale because of the corrugated nature of the probe. Namely, plasmon resonances centered at or around the different nonlinear signals are tracked through TPPL, and they are found to selectively enhance nonlinear signals with closely matched optical resonances.

Femtosecond (fs)-resolved photoemission electron micrographs of single and few-layer MoSe$_2$ track exciton dynamics in this model 2D quantum material system with joint nanometer (nm)-fs resolution. We illustrate the latter using two-photon photoemission following 400 nm fs irradiation to probe and tunable near-IR excitation (770-840 nm) to populate the lowest accessible A-exciton states in MoSe$_2$. Through distinct imaging contrasts in single/few-layer regions, we spatio-temporally visualize exciton lifetimes in multilayers that are ~24 times longer than their monolayer analogs. We also find that tuning the near-IR excitation source allows us to track exciton resonances that vary on the nm scale. In effect, our approach allows us to track exciton resonances and dynamics in real-space and in real-time, which is not currently possible (or at least trivially implemented) using conventional all-optical, nano-optical, and all-electron-based microscopies.

Background Small for gestational age (SGA) birthweight, a risk factor of infant mortality and delayed child development, is associated with maternal educational attainment. Maternal tobacco smoking during pregnancy could contribute to this association. We aimed to quantify the contribution of maternal smoking during pregnancy to social inequalities in child birthweight for gestational age (GA). Methods Data come from the French nation-wide ELFE cohort study, which included 17,155 singletons. Birthweights for GA were calculated using z-scores. Associations between maternal educational attainment, tobacco smoking during pregnancy and child birthweight for GA were ascertained using mediation analysis. Mediation analyses were also stratified by maternal pre-pregnancy body mass index.Results Low maternal educational attainment was associated with an increased odd of tobacco smoking during pregnancy (adjusted OR (ORa)=2.58 [95% CI 2.34, 2.84]) as well as a decrease in child birthweight for GA (RRa=0.94 [95% 0.91, 0.98]). Tobacco smoking during pregnancy was associated with a decrease in offspring birthweight for GA (RRa=0.73 [95% CI 0.70, 0.76]). Mediation analysis suggests that 39% of the effect of low maternal educational attainment on offspring birthweight for GA was mediated by smoking during pregnancy. A more important direct effect of maternal educational attainment on child birthweight for GA was observed among underweight women (RRa=0.82 [95%CI 0.72, 0.93]).Conclusions The relationship between maternal educational attainment and child birthweight for GA is strongly mediated by smoking during pregnancy. Reducing maternal smoking could lessen the occurrence of infant SGA and decrease socioeconomic inequalities in birthweight for GA.Keywords Birthweight, educational attainment, tobacco smoking, pregnancy, mediation analysis, health inequalities

Background An advantaged socioeconomic position (SEP) and satisfying social support during pregnancy (SSP) have been found to be protective factors of maternal postpartum depression (PDD). An advantaged SEP is also associated with satisfying SSP, making SSP a potential mediator of social inequalities in PPD. SEP, SSP and PPD are associated with migrant status. The aim of this study was to quantify the mediating role of SSP in social inequalities in PPD regarding mother's migrant status. Methods A sub-sample of 15,000 mothers from the French nationally-representative ELFE cohort study was used for the present analyses. SEP was constructed as a latent variable measured with educational attainment, occupational grade, employment, financial difficulties and household income. SSP was characterized as perceived support from partner (good relation, satisfying support and paternal leave) and actual support from midwives (psychosocial risk factors assessment and antenatal education). Mediation analyses with multiple mediators, stratified by migrant status were conducted. Results Study population included 76% of non-migrant women, 12% of second and 12% of first generation migrant. SEP was positively associated with support from partner, regardless of migrant status. Satisfying partner support was associated with a 8 (non-migrant women) to 11% (first generation migrant women) reduction in PPD score. Limitations History of depression was not reported.Conclusions Partner support could reduce social inequalities in PPD. This work supports the need of interventions, longitudinal and qualitative studies including fathers and adapted to women at risk of PPD to better understand the role of SSP in social inequalities in PPD. Keywords social support, postpartum depression, epidemiology, social inequalities, pregnancy, mediation analysis

We describe a simplified approach to simulating Raman spectra using ab initio molecular dynamics (AIMD) calculations. Our protocol relies on on-the-fly calculations of approximate molecular polarizabilities using a sum over orbitals (as opposed to states) method.

Plasmon-enhanced chemical transformations at the solid-liquid interface can be imaged with high sensitivity, chemical selectivity, and nanoscale precision through tip-enhanced Raman scattering (TERS). We demonstrate the latter for the first time through measurements aimed at (i) locating plasmonic hotspots at the solid-liquid interface at which chemical transformations take place, (ii) monitoring the evolution from reactants to products through their distinct Raman spectra, and (iii) 2D correlation analysis of Raman time trajectories to unambiguously extract the spectral components that mark chemical transformation and to understand the correlations between the product and parent signatures. For our proof-of-principle study, we select a model plasmon-enhanced chemical process, namely, the dimerization of p-nitrothiophenol to dimercaptoazobenzene, but now, at the solid-liquid interface. Our plasmonic construct otherwise consists of chemically functionalized gold microplates in aqueous solution, which we image using a gold-coated TERS probe irradiated at 633-nm. Overall, we demonstrate chemical reaction imaging in aqueous solution via TERS for the first time, herein, at a pixel-limited lateral spatial resolution of 10 nm.

We demonstrate fast (0.25 s/pixel) nanoscale chemical imaging in aqueous solution via tip-enhanced Raman scattering (TERS), with sub-15 nm spatial resolution. The substrate consists of 4-mercaptobenzonitrile-functionalized monocrystalline gold triangular platelets immersed in H2O, which we map using a gold-coated AFM probe illuminated using a 633 nm laser source. We find that the recorded TERS images trace the enhanced local optical fields that are sustained towards the edges and corners of the triangles. In effect, we directly map local optical fields of a plasmonic substrate in aqueous solution through molecular TERS. Overall, our described platform and approach may generally be used for chemical and biological imaging, and potentially, to follow chemical transformations at solid-liquid interfaces through TERS.

The development of increasingly complex IoT systems requires large engineering environments. These environments generally consist of tools from different vendors and are not necessarily integrated well with each other. In order to automate various analyses, queries across resources from multiple tools have to be executed in parallel to the engineering activities. In this paper, we identify the necessary requirements on such a query capability and evaluate different architectures according to these requirements. We propose an improved lifecycle query architecture, which builds upon the existing Tracked Resource Set (TRS) protocol, and complements it with the MQTT messaging protocol in order to allow the data in the warehouse to be kept updated in real-time. As part of the case study focusing on the development of an IoT automated warehouse, this architecture was implemented for a toolchain integrated using RESTful microservices and linked data., Comment: 12 pages, workshop

Introduction: As humankind ventures further into the depths of space, planning is already underway for long-duration exploration missions that will test the bounds of human performance. Deep space travel will include added risk related to stressors from the isolated, confined, and extreme (ICE) environment that lies outside the boundaries of low-earth orbit. Currently, selective serotonin reuptake inhibitors (SSRIs) are considered the gold standard treatment for many mental health diagnoses including anxiety and depression; however, SSRIs are also associated with several undesired side effects. The utility of non-pharmacologic therapies for the management of behavioral health conditions has not yet been fully explored. Methods: A comprehensive literature search was performed using PubMed and relevant articles pertaining to the psychological impacts of ICE environments, use of SSRIs in spaceflight, side effects associated with SSRIs, and non-pharmacologic treatments for anxiety and depression were reviewed. Over 100 studies were reviewed in total. Results: Reduced bone mineral density, impaired hemostatic function, significant individual variability resulting from gene polymorphisms, and drug-drug interactions are well described adverse effects of SSRIs that may complicate their operational use in the deep space environment. Many non-pharmacologic therapies have been utilized with success for the management of behavioral health conditions in the terrestrial environment that may show promise for long duration missions. Discussion: Although SSRIs have long been considered standard of care treatment for many behavioral health conditions, we cannot trivialize the risk that prolonged pharmacologic therapy may pose. The need to mitigate these risks by exploring alternative non-pharmacologic therapies has never been more relevant.

Non-resonant tip-enhanced Raman images of dimercaptostilbene on silver reveal that different vibrational resonances of the reporter are selectively enhanced at different sites on the metal substrate. Sequentially recorded images track molecular diffusion within the diffraction-limited laser spot which illuminates the substrate. In effect, the recorded time resolved (dt = 10 s) pixelated images (25 nm x 8 cm-1) broadcast molecule-local field interactions which take place on much finer scales., Comment: 16 pages, 3 figures

The Pilot-Mission Exploitation Platform (Pi-MEP) for salinity is an ESA initiative originally meant to support and widen the uptake of Soil Moisture and Ocean Salinity (SMOS) mission data over the ocean. Starting in 2017, the project aims at setting up a computational web-based platform focusing on satellite sea surface salinity data, supporting studies on enhanced validation and scientific process over the ocean. It has been designed in close collaboration with a dedicated science advisory group in order to achieve three main objectives: gathering all the data required to exploit satellite sea surface salinity data, systematically producing a wide range of metrics for comparing and monitoring sea surface salinity products’ quality, and providing user-friendly tools to explore, visualize and exploit both the collected products and the results of the automated analyses. The Salinity Pi-MEP is becoming a reference hub for the validation of satellite sea surface salinity missions by providing valuable information on satellite products (SMOS, Aquarius, SMAP), an extensive in situ database (e.g., Argo, thermosalinographs, moorings, drifters) and additional thematic datasets (precipitation, evaporation, currents, sea level anomalies, sea surface temperature, etc.). Co-localized databases between satellite products and in situ datasets are systematically generated together with validation analysis reports for 30 predefined regions. The data and reports are made fully accessible through the web interface of the platform. The datasets, validation metrics and tools (automatic, user-driven) of the platform are described in detail in this paper. Several dedicated scientific case studies involving satellite SSS data are also systematically monitored by the platform, including major river plumes, mesoscale signatures in boundary currents, high latitudes, semi-enclosed seas, and the high-precipitation region of the eastern tropical Pacific. Since 2019, a partnership in the Salinity Pi-MEP project has been agreed between ESA and NASA to enlarge focus to encompass the entire set of satellite salinity sensors. The two agencies are now working together to widen the platform features on several technical aspects, such as triple-collocation software implementation, additional match-up collocation criteria and sustained exploitation of data from the SPURS campaigns

With a Green's function formalism we obtain the eigenvalue spectrum of a tight-binding one-dimensional exciton model characterized by a contact interaction, a Coulombic electron and hole attraction, the Heller-Marcus exciton-hopping energy and an external constant and homogeneous electric field. The resulting eigenvalue spectrum, in the form of an unevenly spaced Wannier-Stark ladder with envelope profiles, is used to obtain the effective mass of the exciton by the application of the Mattis-Gallinar effective mass formula [D. C. Mattis and J.-P. Gallinar, {\it Phys. Rev. Lett.} {\bf{53}}, 1391 (1984)]. We obtain positive and negative effective masses for the exciton. The inverse effective mass may oscillate periodically as a function of the inverse of the electric field, with the frequency of oscillation linearly dependent upon the tight-binding hopping matrix element. The exciton radius is also obtained with the Green's function formalism, and it too exhibits Keldysh-like field dependent oscillations, as well as abrupt variations associated to strongly avoided crossings in the eigenvalue spectrum. Finally, some comments are made about the experimental relevance of our results., Comment: Latex 7 pages (RevTeX4), 9 eps figures