Search for: All records

The ability to compare objects, scenes, or situations is crucial for effective decision-making and problem-solving in everyday life. For instance, comparing the freshness of apples enables better choices during grocery shopping, while comparing sofa designs helps optimize the aesthetics of our living space. Despite its significance, the comparative capability is largely unexplored in artificial general intelligence (AGI). In this paper, we introduce MLLM-COMPBENCH, a benchmark designed to evaluate the comparative reasoning capability of multimodal large language models (MLLMs). MLLM-COMPBENCH mines and pairs images through visually oriented questions covering eight dimensions of relative comparison: visual attribute, existence, state, emotion, temporality, spatiality, quantity, and quality. We curate a collection of around 40K image pairs using metadata from diverse vision datasets and CLIP similarity scores. These image pairs span a broad array of visual domains, including animals, fashion, sports, and both outdoor and indoor scenes. The questions are carefully crafted to discern relative characteristics between two images and are labeled by human annotators for accuracy and relevance. We use MLLM-COMPBENCH to evaluate recent MLLMs, including GPT-4V(ision), Gemini-Pro, and LLaVA-1.6. Our results reveal notable shortcomings in their comparative abilities. We believe MLLM-COMPBENCH not only sheds light on these limitations but also establishes a solid foundation for future enhancements in the comparative capability of MLLMs. 

Abstract Reliable subseasonal-to-seasonal (S2S) precipitation prediction is highly desired due to the great socioeconomical implications, yet it remains one of the most challenging topics in the weather/climate prediction research area. As part of the Impact of Initialized Land Temperature and Snowpack on Sub-seasonal to Seasonal Prediction (LS4P) project of the Global Energy and Water Exchanges (GEWEX) program, twenty-one climate models follow the LS4P protocol to quantify the impact of the Tibetan Plateau (TP) land surface temperature/subsurface temperature (LST/SUBT) springtime anomalies on the global summertime precipitation. We find that nudging towards reanalysis winds is crucial for climate models to generate atmosphere and land surface initial conditions close to observations, which is necessary for meaningful S2S applications. Simulations with nudged initial conditions can better capture the summer precipitation responses to the imposed TP LST/SUBT spring anomalies at hotspot regions all over the world. Further analyses show that the enhanced S2S prediction skill is largely attributable to the substantially improved initialization of the Tibetan Plateau-Rocky Mountain Circumglobal (TRC) wave train pattern in the atmosphere. This study highlights the important role that initial condition plays in the S2S prediction and suggests that data assimilation technique (e.g., nudging) should be adopted to initialize climate models to improve their S2S prediction. 

Abstract. Plant and microbial nitrogen (N) dynamics and N availability regulate the photosynthetic capacity and capture, allocation, and turnover of carbon (C) in terrestrial ecosystems. Studies have shown that a wide divergence in representations of N dynamics in land surface models leads to large uncertainties in the biogeochemical cycle of terrestrial ecosystems and then in climate simulations as well as the projections of future trajectories. In this study, a plant C–N interface coupling framework is developed and implemented in a coupled biophysical-ecosystem–biogeochemical model (SSiB5/TRIFFID/DayCent-SOM v1.0). The main concept and structure of this plant C–N framework and its coupling strategy are presented in this study. This framework takes more plant N-related processes into account. The dynamic C/N ratio (CNR) for each plant functional type (PFT) is introduced to consider plant resistance and adaptation to N availability to better evaluate the plant response to N limitation. Furthermore, when available N is less than plant N demand, plant growth is restricted by a lower maximum carboxylation capacity of RuBisCO (Vc,max), reducing gross primary productivity (GPP). In addition, a module for plant respiration rates is introduced by adjusting the respiration with different rates for different plant components at the same N concentration. Since insufficient N can potentially give rise to lags in plant phenology, the phenological scheme is also adjusted in response to N availability. All these considerations ensure a more comprehensive incorporation of N regulations to plant growth and C cycling. This new approach has been tested systematically to assess the effects of this coupling framework and N limitation on the terrestrial carbon cycle. Long-term measurements from flux tower sites with different PFTs and global satellite-derived products are employed as references to assess these effects. The results show a general improvement with the new plant C–N coupling framework, with more consistent emergent properties, such as GPP and leaf area index (LAI), compared to the observations. The main improvements occur in tropical Africa and boreal regions, accompanied by a decrease in the bias in global GPP and LAI by 16.3 % and 27.1 %, respectively. 

Abstract NONEXPRESSER OF PATHOGENESIS-RELATED GENES 1 (NPR1) is the master regulator of salicylic acid-mediated basal and systemic acquired resistance in plants. Here, we report that NPR1 plays a pivotal role in restricting compatible infection by turnip mosaic virus, a member of the largest plant RNA virus genus Potyvirus , and that such resistance is counteracted by NUCLEAR INCLUSION B (NIb), the viral RNA-dependent RNA polymerase. We demonstrate that NIb binds to the SUMO-interacting motif 3 (SIM3) of NPR1 to prevent SUMO3 interaction and sumoylation, while sumoylation of NIb by SUMO3 is not essential but can intensify the NIb–NPR1 interaction. We discover that the interaction also impedes the phosphorylation of NPR1 at Ser11/Ser15. Moreover, we show that targeting NPR1 SIM3 is a conserved ability of NIb from diverse potyviruses. These data reveal a molecular “arms race” by which potyviruses deploy NIb to suppress NPR1-mediated resistance through disrupting NPR1 sumoylation. 

The incessant mutations of viruses, variable immune responses, and likely emergence of new viral threats necessitate multiple approaches to novel antiviral therapeutics. Furthermore, the new antiviral agents should have broad-spectrum activity and be environmentally stable. Here, we show that biocompatible tapered CuS nanoparticles (NPs) efficiently agglutinate coronaviruses with binding affinity dependent on the chirality of surface ligands and particle shape.L-penicillamine-stabilized NPs with left-handed curved apexes display half-maximal inhibitory concentrations (IC₅₀) as low as 0.66 pM (1.4 ng/mL) and 0.57 pM (1.2 ng/mL) for pseudo-type SARS-CoV-2 viruses and wild-type Wuhan-1 SARS-CoV-2 viruses, respectively, which are about 1,100 times lower than those for antibodies (0.73 nM). Benefiting from strong NPs–protein interactions, the same particles are also effective against other strains of coronaviruses, such as HCoV-HKU1, HCoV-OC43, HCoV-NL63, and SARS-CoV-2 Omicron variants with IC₅₀values below 10 pM (21.8 ng/mL). Considering rapid response to outbreaks, exposure to elevated temperatures causes no change in the antiviral activity of NPs while antibodies are completely deactivated. Testing in mice indicates that the chirality-optimized NPs can serve as thermally stable analogs of antiviral biologics complementing the current spectrum of treatments. 

Stabilizing perovskite solar cells requires consideration of all defective sites in the devices. Substantial efforts have been devoted to interfaces, while stabilization of grain boundaries received less attention. Here, we report on a molecule tributyl(methyl)phosphonium iodide (TPI), which can convert perovskite into a wide bandgap one-dimensional (1D) perovskite that is mechanically robust and water insoluble. Mixing TPI with perovskite precursor results in a wrapping of perovskite grains with both grain surfaces and grain boundaries converted into several nanometer-thick 1D perovskites during the grain formation process as observed by direct mapping. The grain wrapping passivates the grain boundaries, enhances their resistance to moisture, and reduces the iodine released during light soaking. The perovskite films with wrapped grains are more stable under heat and light. The best device with wrapped grains maintained 92.2% of its highest efficiency after light soaking under 1-sun illumination for 1900 hours at 55°C open-circuit condition.