Integrated satellite-terrestrial networks (ISTNs) provide crucial support for ubiquitous connectivity, whereas the inherent openness poses significant security threats. Covert communication that achieves intrinsic protection by concealing transmission behavior has become an ideal solution. This work studies a downlink covert satellite communication scenario under signal overlay. Existing research mostly carries out optimizations in the power domain. However, since waveforms serve as the physical carriers of information, their design critically impacts the covert communication performance via different adjustability to channel state information (CSI). Specifically, blindly enhancing demodulation reliability does not necessarily lead to better covert capacity and may even be counterproductive. OFDM is vulnerable to carrier frequency offset, while OTFS lacks the flexibility to accommodate CSI, resulting in difficulties applying existing candidate waveforms. In light of this, we propose a nimble waveform named orthogonal fractional dual index multiplexing (OFDIM) tolerant of fractional delay and Doppler. It enables common modulation schemes through integer indices and creates a fractional domain resilient to CSI. The closed-forms of the maximum covert transmit power, reliability, and spectral efficiency are derived. Fractional indices are optimized to improve the performance based on CSI. Simulations validate the theoretical analysis and demonstrate the benefits of the fractional domain in covert communication. Additionally, OFDIM incorporates a complementary security mechanism where the indices can act as dynamic ciphers and guarantee certain security even if the communication behavior is exposed.

The sixth generation mobile communication system promises to achieve ubiquitous coverage. Although integrated satellite-terrestrial networks (ISTNs) offer a direct resolvent, high mobility and openness pose challenges to reliability. Orthogonal time frequency space (OTFS) mitigates the limitation of orthogonal frequency division multiplexing (OFDM) in combating time-selective fading with the cost of additional complexity. Nonetheless, the diverse scenarios lead to differentiation in requirements from the perspective of waveform design. Therefore, a single waveform is inadequate to be uniformly adopted in ISTNs, further aggravated by the difficulty in upgrading satellite hardware. Consequently, there is an urgent need for a soft approach capable of switching between traditional waveforms while maintaining reliable communication under adverse conditions. A fractional dual index transform is introduced, enabling seamless switching and fusion across time, frequency, delay, and Doppler domains through two tunable indices. Based on this, fractional dual index division multiplexing (FDIDM) is designed along with input-output relationships and feasible decoders. FDIDM generalizes typical waveforms like OFDM and OTFS as special cases with the same complexity and improves performance through indices optimization at the cost of additional computational overhead. Theoretical analysis is conducted to obtain the closed-form symbol error rate (SER) and performance boundaries of FDIDM. Simulation results validate the deduction and demonstrate the advantages of cross-domain modulation over existing waveforms.

Attention is the cornerstone of modern Large Language Models (LLMs). Yet its quadratic complexity hinders efficiency and scalability, especially for long-context processing. A promising approach is to leverage sparsity in attention. However, existing sparsity-based solutions predominantly rely on predefined patterns or heuristics at the attention head level, struggling to adapt dynamically to different contexts efficiently. We propose SeerAttention, a simple yet effective attention mechanism that directly learns the block-level attention sparsity from the LLM itself. Inspired by the gating mechanism in Mixture of Experts (MoE), SeerAttention augments the conventional attention with a **learnable gate** that **selectively activates important blocks** within the attention map. Specifically, the gate first pools the query (Q) and key (K) tensors along the sequence dimension and processes them through learnable linear layers. The resulting matrices are then multiplied together to produce the gating scores, which are used to predict block-level attention sparsity. Combined with our block-sparse FlashAttention kernel, SeerAttention can achieve significant speedup on GPUs. When applied to pre-trained LLMs, SeerAttention only requires training the gate parameters in a lightweight self-distillation manner, allowing rapid convergence. Our evaluation results demonstrate that SeerAttention achieves better model accuracy and lower latency for long-context pre-filling compared to prior methods. Code is available at: https://github.com/microsoft/SeerAttention.

Real-time precise point positioning (PPP) is an innovative and useful atmosphere remote sensing tool to estimate zenith tropospheric delays (ZTDs) to support various emerging meteorological applications such as weather and climate research. However, additional costs on internet accesses or service subscription licenses are required to obtain real-time precise satellite orbit and clock products from external PPP augmentation service operators. We present the real-time precise estimation of ZTD with freely and openly accessible GNSS satellite-based PPP augmentation services, i.e., BeiDou navigation satellite system (BDS) PPP-B2b service, Galileo high accuracy service (HAS), and quasi-zenith satellite system (QZSS) Multi-GNSS ADvanced Orbit and Clock Augmentation-PPP (MADOCA-PPP) service. GNSS observations of 49 days from 15 stations in the Asia-Pacific region are used for performance assessment and comparison. It shows that the real-time PPP estimated ZTDs agree well with International GNSS Service (IGS) final ZTDs, with mean standard deviation (STD) values of 20.0, 17.5, and 9.5 mm for BDS PPP-B2b, Galileo HAS, and QZSS MADOCA-PPP, respectively. The derived real-time zenith wet delays (ZWDs) are further compared with those derived from Integrated Global Radiosonde Archive (IGRA) and the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) datasets, which confirms that the real-time PPP estimated ZWDs are accurate with mean STDs of smaller than 30 mm. It concludes that real-time PPP with GNSS satellite-based real-time PPP augmentation correction services will be a flexible and cost-effective tool to derive real-time precise ZTDs/ZWDs with high spatiotemporal resolution in all-weather conditions.

Peritoneal dialysis (PD) removes waste products from blood when the kidney is malfunctioned. Since there is no clear criterion for PD recommendation for patients with kidney disease, existing machine learning models (ML), which rely on credible decision criterion, are ineffective in making PD eligibility decisions, especially when the correlated traits or indicators (patterns) inherent in the PD data are diverse and subtle. Furthermore, the lack of interpretable transparency in traditional ML also weakens the credibility of the decision they produce. Hence, an in-depth knowledge of the patients' characteristics is needed to render a clearer picture of the decision-making process and model to detect the rare PD eligibility cases. In this paper, we extend our previous work (Attribute-Value-Association Discovery and Disentanglement (ADD)), to an extended ADD for PD data analysis (PD-ADD) to overcome these problems. We show that PD-ADD is able to discover association patterns of patient profiles and symptoms to reveal PD characteristics and detect eligible rare cases. Experimental results show that PDADD is much superior to existing unsupervised clustering (with accuracy of 89.87% vs 73.37% of K-Means). It also enables straightforward interpretation of the underlying relations of patient characteristics in an unsupervised setting.