Modern large language model (LLM) serving systems confront inefficient GPU utilization due to the fundamental mismatch between compute-intensive prefill phase and memory-bound decode phase. While current practices attempt to address this by organizing these phases into hybrid batches, such solutions create an inefficient tradeoff that sacrifices either throughput or latency, leaving substantial GPU resources underutilized. For this, we identify two key root causes: 1) the prefill phase suffers from suboptimal compute utilization due to wave quantization and attention bottlenecks, and 2) hybrid batching disproportionately prioritizes latency over throughput, wasting both compute resources and memory bandwidth. To mitigate the issues, we present Bullet, a novel spatial-temporal orchestration system that eliminates these inefficiencies through fine-grained phase coordination. Bullet enables concurrent execution of prefill and decode requests, while dynamically provisioning GPU resources based on real-time performance modeling. By integrating SLO-aware scheduling and adaptive resource allocation, Bullet maximizes GPU utilization without compromising latency targets. Experimental evaluations on real-world workloads demonstrate that Bullet delivers 1.26× average throughput gains (up to 1.55×) over state-of-the-arts, while consistently meeting latency constraints.

To accelerate large language model (LLM) inference, pipeline parallelism partitions model layers into sequential stages, each assigned to a different device for concurrent execution. However, this method often suffers from pipeline bubbles caused by imbalanced computation in the tail stage. While upstream stages focus solely on layer-forward operations, the final stage must also handle post-processing tasks like sampling, introducing significant latency. This uneven workload leads to pipeline misalignment, forcing upstream stages to idle and degrading overall performance. Existing frameworks typically distribute layers evenly across stages without accounting for computational load differences. To address this, we propose DynaPipe, a dynamic layer redistribution scheme that adaptively balances computation by predicting execution latency in real time. Moreover, we introduce an asynchronous key-value (KV) cache migration coordinator to enable non-blocking layer redistribution during inference. Experiments on representative LLMs demonstrate that DynaPipe reduces average end-to-end request latency by 8% to 49% across diverse workloads, outperforming state-of-the-art pipeline parallelism systems.

The bidirectional reflectance factor (BRF) is a key parameter for understanding the radiative transfer process within cities, influenced by the sun-target-sensor geometry. Simulating urban BRF poses significant challenges due to the complex three-dimension (3-D) structures of urban landscapes. Previous attempts have been facing significant hurdles, either inadequately addressing the complexities of real-world urban structures or consuming too many computing resources. To overcome these challenges, this study introduces an analytical model—urban bidirectional reflectance analytical model (UBRAM), which takes advantage of landscape-level geometric and radiometric parameters incorporating 3-D urban structures to simulate the BRF of neighborhood-scale urban landscapes ($100\times 100- 1000\times 1000$m). UBRAM underwent verification using simulations from the well-known radiative transfer model (discrete anisotropic radiative transfer, DART) under various solar-illumination geometries and diverse urban morphologies. Results demonstrated a good agreement between UBRAM and DART, with an overall${R} ^{2}$ranging from 0.829 to 0.983 and the shape similarity index exceeding 0.99 in most scenes, underscoring the effectiveness of UBRAM. Notably, leveraging a modular design, UBRAM enables independent processing of 3-D urban scenes and radiative transfer simulation, ensuring efficiency and scale applicability. This study establishes UBRAM as a viable model, offering a novel approach for quantifying BRF of neighborhood-scale urban surfaces. Moreover, UBRAM requires a limited number of easily obtainable parameters, facilitating its straightforward adaptation for application in other urban areas. UBRAM facilitates the simulation of radiative transfer processes based on real-world 3-D urban models, thereby enhancing the accuracy of scene rendering and downstream applications such as urban heat islands and environment monitoring.

Lithium-ion batteries, which are vital for powering mobile devices, experience performance degradation over time due to capacity fading and other aging phenomena, thereby presenting safety risks. This work introduces the DeTransformer-Physics model, a predictive model for battery capacity. This model synergistically integrates physical knowledge and a de-noising autoencoder to enhance predictive accuracy. Firstly, it employs a denoising autoencoder as a preprocessing step to reconstruct denoised input data, thereby facilitating the extraction of more effective input data. Subsequently, a model describing the procedure for lithium-ion battery capacity degradation is derived and incorporated into a neural network, with the model training constrained by physical principles. Finally, using the lithium-ion battery degradation dataset from NASA, we demonstrate that incorporating the denoising autoencoder and embedding physical knowledge substantially improves the predictive accuracy of the model.

Underwater object detection has attracted increasing interest for its wide application in various underwater tasks. However, due to underwater image quality degradation and the lack of large-scale underwater object datasets, many underwater detectors suffer from low detection performance. To address the issues, we not only propose a novel underwater transformer detector with multi-scale feature enhancement and query optimization, named QO-Net, but also construct a new underwater object detection dataset, called UODD. Specifically, a Conv-Trans Layer is developed as the unit of QO-Net, which effectively learns multi-scale image feature representation through CNN and simultaneously captures the dependencies among different positions in the sequence data through Transformer, enabling QO-Net to process underwater image sequence information over longer distances. An effective combination can enhance the representation of multi-scale features. Then, QO-Net develops a positional query enhancement strategy to optimize the spatial prior of positional queries, thereby speeding up the convergence of the network training. In addition, UODD also contains more than 20,000 underwater images for training and validation, with a variety of rich underwater categories. Extensive experiments on UODD, Brackish, and TrashCan datasets demonstrate that QO-Net presents favorable detection performance against state-of-the-art methods in terms of robustness and accuracy.

3D single object tracking (SOT) based on point cloud has attracted much attention due to its important role in machine vision and autonomous driving. Recently, M2-Track proposes a two-stage tracking structure centered on motion, but they ignore the effect of segmentation errors in sparse point cloud scenarios, which hinder the ability of networks to accurately represent tracking targets. To solve the problems, we propose an efficient 3D single object tracker (Abbr. EST) that can effectively segment point cloud features. Firstly, the proposed fusion segmentation module makes up for the feature loss caused by the downsampling strategy and enhances the ability of the network to recognize foreground points. In addition, the global embedded module is used to further focus on the crucial features of the target. This module provides global information by using residual networks and adding background information. Numerous experiments conducted on KITTI and NuScenes benchmarks show that EST achieves superior point cloud tracking in both performance and efficiency.

Three-dimensional (3D) urban data opens up new possibilities and challenges for urban research. In this paper, we present an analytical model for urban bidirectional reflectance distribution function based on geometric parameters extracted from real-world 3D data. The analytical model relies on the simplified urban scenes, which have the same geometric parameters as real urban 3D scenes. The bidirectional reflectance factor of the 4 urban scenes in the two major cities of Beijing and Shanghai in China is simulated with the urban analytical model. The innovations of this paper are as follows: 1) a simple and efficient analytical model is developed, considering the complexity of urban structure and the optical properties of ground objects, and 2) the analytical model is backed up by real 3D data and can be applied in various forms of urban scenes.

In this study, along-scan performance of FengYun-3C microwave radiation imager (MWRI) is examined utilizing one-year data over ocean. Generally-good uniformity along azimuth scan position is found with a TB bias less than 1 K at the edge-of-scan. However, discernible errors can be still found and the behaviors of the scan-position-dependent bias are different for ascending and descending phases of the orbits. It is thus suggested that descending and ascending passes should be taken into account, separately, when correcting antenna pattern in calibration processes.

The Soil Moisture Experiment in the Luan River (SMELR) in 2018 was carried out toward developing of new satellite mission opportunities in China. It was designed to explore several scientific and technical questions regarding to soil moisture remote sensing and its application. Various passive, active microwave and optical observations were collected by both airborne and satellite platforms. Ground-sampling of soil moisture/temperature, vegetation and roughness were conducted close in time to the airborne acquisitions at 200 to 2000 m scales. Ground-based measurements of microwave emission and scattering, emissivity and reflectance spectra, evapotranspiration and radiation were conducted along the flight areas. Moreover, two in-situ networks that cover the Shandian (100×100 km) and Xiaoluan (25×25 km) river basins were established to provide continuous measurements of soil moisture and temperature profiles (3-50 cm). This paper describes the overview of the experimental design, obtained data sets and initial results.

Microwave radiation imager (MWRI) is regarded as one of the most important microwave payloads on board China FengYun-3C Meteorological Satellite. The instrument suffers from calibration anomalies and exhibits observation- background (O-B) calibration bias difference between the ascending and descending passes at all channels (hereinafter AD bias). The calibration bias difference of MWRI between ascending and descending orbits hampers data assimilation in the numerical weather predictions and reanalysis systems. This paper proposes a physical-based correction algorithm for MWRI calibration, following a brief introduction to the calibration process of the instrument. The relationship between the observed brightness temperatures and the physical temperature of the hot load reflector is established to mitigate the intrusion of the emissive hot reflector at all channels which was not accurately estimated in the previous calibration process. Before- and after-correction comparisons using one-year observations show that the AD bias is effectively reduced, i.e., from ~2 K before correction to less than 0.2 K after correction, when rectifying the emissivity of the hot reflector in the calibration equation, whereas the change in the mean values of MWRI radiance is negligible.

This paper analyzes the effects of cloud supercooled liquid water (SCLW) on the polarization difference (PD) and brightness temperature (TB) generated by horizontally oriented snow particles at microwave band. Radiative transfer (RT) calculations from six realistic snowfall profiles selected from European Centre for Medium-Range Weather Forecasts (ECMWF) datasets indicate, that the existence of SCLW has noticeable impact on PD and TB at three window frequencies (150 GHz, 243 GHz and 664 GHz) for, respectively, light, median and heavy snowfall in dry and wet air conditions. It is implied that accurate information on liquid water is required to interpret polarimetric observations of ice clouds and snowfall for future satellite missions.

Microwave Radiation Imager (MWRI) onboard the FengYun (FYJ-3A/B satellites observes the Earth atmosphere at 10.65, 18.7, 23.8, 36.5 and 89.0 GHz with each having dual polarization. Its calibration system is uniquely designed with a main reflector viewing both cold and hot calibration targets. Two quasi-optical reflectors are used to reflect the radiation from hot load and cold space to the main reflector. Soon after FY3b was successfully launched in November 2010, evaluation of on-orbit calibration accuracy of MWRI was carried out. In this paper, CRTM was used to simulate the MWRI observations over ocean, by using GDAS data as model inputs. "O-B" Results and "Double difference " results show that: 1). the on-orbit calibration status of MWRI is stable, 2). the brightness temperatures from MWRI observation are highly consistent with those derived from AMSR-E and model simulation.

The Microwave Radiation Imager (MWRI) on board the FengYun-3A/B satellites observes the Earth atmosphere at 10.65, 18.7, 23.8, 36.5, and 89.0 GHz with each having dual polarization. Its calibration system is uniquely designed with a main reflector viewing both cold and hot calibration targets. Two quasi-optical reflectors are used to reflect the radiation from the hot load and cold space to the main reflector. In the MWRI calibration process, a radiation loss in the beam transmission path must be taken into account. The loss factor in the hot load transmission path is derived using the antenna pattern data measured on ground and satellite data observing over the Amazon forest where the scene temperature is steady and close to the hot load. The instrument nonlinearity factors at different channels are also evaluated over a wide range of brightness temperatures and compared with the results from the ground vacuum test. After a cross-calibration with Windsat data, atmospheric products are derived from MWRI brightness temperatures with the accuracy similar to those from the legacy sensors (e.g., the Special Sensor Microwave/Imager).