Explore the vast range of titles available.

We introduce Gemini Embedding 2, a native multimodal embedding model that allows embedding video, audio, image, and text modalities in a unified representation space. We leverage the multimodal capabilities of Gemini to produce embeddings for arbitrary combinations of interleaved inputs across all these modalities that generalize well across a wide variety of tasks. Applying large-scale contrastive learning in a multi-task multi-stage training setup, we achieve state-of-the-art performance on key embedding benchmarks including unimodal, cross-modal, and multimodal retrieval spanning a diverse set of tasks. We show that our embedding model demonstrates strong performance (with a score of 62.9 R@1 on MSCOCO, 68.8 NDCG@10 on Vatex, 69.9 on MTEB multilingual and 84.0 on MTEB Code) across a variety of tasks surpassing the performance of specialized models. These unified capabilities make Gemini Embedding 2 a promising candidate for downstream use cases such as RAG, recommendation and search. Furthermore, its robust zero-shot performance across distinct fields - from astronomy and bioscience to fine arts and the culinary arts - establishes it as a highly reliable, out-of-the-box representation even for specialized domains.

We introduce EmbeddingGemma, a new lightweight, open text embedding model based on the Gemma 3 language model family. Our innovative training recipe strategically captures knowledge from larger models via encoder-decoder initialization and geometric embedding distillation. We improve model robustness and expressiveness with a spread-out regularizer, and ensure generalizability by merging checkpoints from varied, optimized mixtures. Evaluated on the Massive Text Embedding Benchmark (MTEB) across multilingual, English, and code domains, EmbeddingGemma (300M) achieves state-of-the-art results. Notably, it outperforms prior top models, both proprietary and open, with fewer than 500M parameters, and provides performance comparable to models double its size, offering an exceptional performance-to-cost ratio. Remarkably, this lead persists when quantizing model weights or truncating embedding outputs. This makes EmbeddingGemma particularly well-suited for low-latency and high-throughput use cases such as on-device applications. We provide ablation studies exploring our key design choices. We release EmbeddingGemma to the community to promote further research.

High quantum efficiency and wide-band detection capability are the major thrusts of infrared sensing technology. However, bulk materials with high efficiency have consistently encountered challenges in integration and operational complexity. Meanwhile, two-dimensional (2D) semimetal materials with unique zero-bandgap structures are constrained by the bottleneck of intrinsic quantum efficiency. Here, we report a near-mid infrared ultra-miniaturized graphene photodetector with configurable 2D potential well. The 2D potential well constructed by dielectric structures can spatially (laterally and vertically) produce a strong trapping force on the photogenerated carriers in graphene and inhibit their recombination, thereby improving the external quantum efficiency (EQE) and photogain of the device with wavelength-immunity, which enable a high responsivity of 0.2 A/W–38 A/W across a broad infrared detection band from 1.55 to 11 µm. Thereafter, a room-temperature detectivity approaching 1 × 10 9  cm Hz 1/2 W −1 is obtained under blackbody radiation. Furthermore, a synergistic effect of electric and light field in the 2D potential well enables high-efficiency polarization-sensitive detection at tunable wavelengths. Our strategy opens up alternative possibilities for easy fabrication, high-performance and multifunctional infrared photodetectors. The integration of 2D materials with metasurfaces can enhance their quantum efficiency, but the approach is usually limited to a narrow spectral band. Here, the authors report the realization of gate-tunable graphene photodetectors combined with all-dielectric periodic slits, leading to enhanced photoresponse in the short-to-long-wave infrared.

Tuneable retarder arrays, such as spatially patterned liquid crystal devices, have given rise to impressive photonic functionality, fuelling diverse applications ranging from microscopy and holography to encryption and communications. Presently these solutions are limited by the controllable degrees of freedom of structured matter, hindering applications that demand photonic systems with high flexibility and reconfigurable topologies. Here we demonstrate a compound modulator that implements a synthetic tuneable arbitrary retarder array as virtual pixels derived by cascading low functionality tuneable devices, realising full dynamic control of its arbitrary elliptical axis geometry, retardance value, and induced phase. Our approach offers unprecedented functionality that is user-defined and possesses high flexibility, allowing our modulator to act as a new beam generator, analyser, and corrector, opening an exciting path to tuneable topologies of light and matter. Retarder arrays enable advanced photonic applications but are limited by controllable flexibility. Here, authors demonstrate a compound modulator that creates synthetic tuneable retarder arrays, offering unprecedented dynamic control of light, enabling new beam generation, analysis, and correction.

Van der Waals coupling is emerging as a powerful method to engineer physical properties of atomically thin two-dimensional materials. In coupled graphene–graphene and graphene–boron nitride layers, interesting physical phenomena ranging from Fermi velocity renormalization to Hofstadter’s butterfly pattern have been demonstrated. Atomically thin transition metal dichalcogenides, another family of two-dimensional-layered semiconductors, can show distinct coupling phenomena. Here we demonstrate the evolution of interlayer coupling with twist angles in as-grown molybdenum disulfide bilayers. We find that the indirect bandgap size varies appreciably with the stacking configuration: it shows the largest redshift for AA- and AB-stacked bilayers, and a significantly smaller but constant redshift for all other twist angles. Our observations, together with ab initio calculations, reveal that this evolution of interlayer coupling originates from the repulsive steric effects that leads to different interlayer separations between the two molybdenum disulfide layers in different stacking configurations. Interlayer coupling between two-dimensional materials is known to result in interesting physical properties. Here, the authors study the effect of a twist angle between two-dimensional molybdenum disulphide on interlayer coupling, observing an indirect bandgap, the size of which depends on the twist angle.

The diamondback moth (DBM), Plutella xylostella (Lepidoptera: Plutellidae) is a very destructive crucifer-specialized pest that has resulted in significant crop losses worldwide. DBM is well attracted to glucosinolates (which act as fingerprints and essential for herbivores in host plant recognition) containing crucifers such as wintercress, Barbarea vulgaris (Brassicaceae) despite poor larval survival on it due to high-to-low concentration of saponins and generally to other plants in the genus Barbarea. B. vulgaris build up resistance against DBM and other herbivorous insects using glucosinulates which are used in plant defense. Aside glucosinolates, Barbarea genus also contains triterpenoid saponins, which are toxic to insects and act as feeding deterrents for plant specialist herbivores (such as DBM). Previous studies have found interesting relationship between the host plant and secondary metabolite contents, which indicate that attraction or resistance to specialist herbivore DBM, is due to higher concentrations of glucosinolates and saponins in younger leaves in contrast to the older leaves of Barbarea genus. As a response to this phenomenon, herbivores as DBM has developed a strategy of defense against these plant biochemicals. Because there is a lack of full knowledge in understanding bioactive molecules (such as saponins) role in plant defense against plant herbivores. Thus, in this review, we discuss the role of secondary plant metabolites in plant defense mechanisms against the specialist herbivores. In the future, trials by plant breeders could aim at transferring these bioactive molecules against herbivore to cash crops.

The blood proteome holds great promise for precision medicine but poses substantial challenges due to the low abundance of most plasma proteins and the vast dynamic range of the plasma proteome. Here we address these challenges with NUcleic acid Linked Immuno-Sandwich Assay (NULISA™), which improves the sensitivity of traditional proximity ligation assays by ~10,000-fold to attomolar level, by suppressing assay background via a dual capture and release mechanism built into oligonucleotide-conjugated antibodies. Highly multiplexed quantification of both low- and high-abundance proteins spanning a wide dynamic range is achieved by attenuating signals from abundant targets with unconjugated antibodies and next-generation sequencing of barcoded reporter DNA. A 200-plex NULISA containing 124 cytokines and chemokines and other proteins demonstrates superior sensitivity to a proximity extension assay in detecting biologically important low-abundance biomarkers in patients with autoimmune diseases and COVID-19. Fully automated NULISA makes broad and in-depth proteomic analysis easily accessible for research and diagnostic applications. Unlocking the blood proteome requires exquisite sensitivity and multiplexing to detect low and high abundance proteins simultaneously. Here the authors describe a 200-plex immunoassay with attomolar sensitivity to detect important low abundance proteins in inflammatory diseases and COVID-19.

Clinical research is necessary for an effective response to an emerging infectious disease outbreak. However, research efforts are often hastily organised and done using various research tools, with the result that pooling data across studies is challenging. In response to the needs of the rapidly evolving COVID-19 outbreak, the Clinical Characterisation and Management Working Group of the WHO Research and Development Blueprint programme, the International Forum for Acute Care Trialists, and the International Severe Acute Respiratory and Emerging Infections Consortium have developed a minimum set of common outcome measures for studies of COVID-19. This set includes three elements: a measure of viral burden (quantitative PCR or cycle threshold), a measure of patient survival (mortality at hospital discharge or at 60 days), and a measure of patient progression through the health-care system by use of the WHO Clinical Progression Scale, which reflects patient trajectory and resource use over the course of clinical illness. We urge investigators to include these key data elements in ongoing and future studies to expedite the pooling of data during this immediate threat, and to hone a tool for future needs.

Plant aquaporins (AQPs) were initially described as a family of membrane-localized proteins exclusively facilitating water transport. Subsequently, sub-sets of plant AQPs have exhibited diverse functionalities beyond water transport. The aquaporin AtPIP2;1, an abundant Plasma membrane Intrinsic Protein in Arabidopsis thaliana , can transport water but also CO 2 , H 2 O 2 and monovalent cations under certain conditions. However, the mechanisms regulating the selectivity of AtPIP2;1, particularly for cations and water, remain to be fully explored. Here we report the outcome of mutating four AtPIP2;1 serine phosphorylation sites to mimic states of phosphorylation and dephosphorylation in loops B and D, and the C-terminal domain. Expression of the mutated proteins in Xenopus laevis oocytes allowed analysis of both water and ion conduction. Concurrent modifications at the four phosphorylation sites may collectively act as a ‘selectivity switch,’ modulating the permeability between cations and water for the homotetramer of AtPIP2;1, allowing for the possibility of simultaneous transport, with one substrate remaining dominant. The reciprocal relationship between cation conductance and water transport fits with the model of a gated ion-permeable pore of the tetramer being dependent on the four individual monomer water conductance states. Notably, in several instances, cation conductance can be turned off, reaching levels comparable to those of the H₂O-injected control, and these instances corresponded with maximal water transport. In contrast, when cation conductance was significantly increased, water transport was reduced but not completely silenced. AtPIP2;1 triple mutant S194A/S280DS283D (A/DD, Loop D and C-terminal regions respectively) displayed very high cation conductance with a selectivity sequence for univalent cations of K +  > Rb +  > Cs +  > Na +  > Li +  > TEA + (tetraethylammonium + ) > choline +  > NMDG + (N-methyl-d-glucamine). In conclusion, our results suggest that post-translational regulations may provide AtPIP2;1 with the flexibility to switch between predominantly cation transport or predominantly water transport. This dynamic ‘switch’ likely contributes to maintaining water and ion homeostasis under diverse environmental conditions.

Hydrologically sensitive areas (HSAs) are runoff-generating areas often targeted for effective water resources planning and management actions. Commonly, HSAs can be mapped as areas in a landscape with a topographic index (TI) greater than a threshold level. This study explored the impact of a gradient of different TI threshold values for delineating HSAs using two popular TIs: a topographic wetness index (TWI) and a soil topographic index (STI). The resultant HSAs for each TI were compared to the Federal Emergency Management Agency (FEMA) 100-year floodplain map in New Jersey and its five water regions. Spatial comparison indicators were used to assess the spatial similarity between the HSAs delineated and the FEMA floodplain map. Such comparisons identified the threshold that delineated HSAs whose spatial distributions were most consistent with the FEMA floodplain at each spatial scale for each TI. For example, the identified threshold for using a TWI to delineate HSAs was 10.5 at the state level; however, this threshold varied by the water region. The HSAs delineated approximate the spatial extent of runoff-contributing areas to the 100-year flood relevant for water resources planning and management actions for flood hazard mitigation.