Background: Serial electrocardiography (ECG) enables detection of subtle cardiac deterioration hours to days before fatal events, yet current clinical practice lacks automated quantification of inter-recording discrepancies, relying on manual comparison instead. Existing deep-learning mortality predictors fail to capture crucial temporal dynamics—either reducing admissions to static probabilities or disregarding death timing. Current methods also inadequately handle irregular ECG intervals, discard temporal context, process ECGs in isolation (losing evolutionary patterns), and exhibit performance degradation across clinical centers due to simplistic handling of non-uniform sampling. We propose a lightweight, plug-and-play Discrepancy-Time Aware Block (DTAB) that can be inserted after CNN feature extractor and before sequential encoder (LSTM/Transformer). DTAB combines two core mechanisms, discrepancy vectors encoding inter-ECG changes and a time-aware gating mechanism for memory decay. The block outputs time-calibrated embeddings that are propagated to a multi-window survival head yielding mortality probabilities for six clinically relevant horizons (24 h–2 months).

Parameter-Efficient Fine-Tuning (PEFT) is a fundamental research problem in computer vision, which aims to tune a few parameters for efficient storage and adaptation of pre-trained vision models. Recently, sensitivity-aware parameter efficient fine-tuning method (SPT) addresses this problem by identifying sensitive parameters and then leveraging its sparse characteristic to combine unstructured and structured tuning for PEFT. However, existing methods only focus on the sparse characteristic of sensitive parameters but overlook its distribution characteristic, which results in additional storage burden and limited performance improvement. In this paper, we find that the distribution of sensitive parameters is not chaotic, but concentrates on a small number of rows or columns in each parameter matrix. Inspired by this fact, we propose a Compact Dynamic-Rank Adaptation-based tuning method for Sensitivity-aware Parameter efficient fine-Tuning, called CDRA-SPT. Specifically, we first identify the sensitive parameters that require tuning for each down-stream task. Then, we reorganize the sensitive parameters by following its row and column into a compact sub-parameter matrix. Finally, a dynamic-rank adaptation is designed and applied at sub-parameter matrix level for PEFT. Its advantage is that the dynamic-rank characteristic of sub-parameter matrix can be fully exploited for PEFT. Extensive experiments show that our method achieves superior performance over previous state-of-the-art methods.

Fuzzing, one of the most popular methods for enhancing software security and quality, relies heavily on the quality of its initial seed corpus to effectively uncover vulnerabilities. Traditional methods for generating initial seed corpora, whether crawl-based or generation-based, struggle with programs that handle highly structured formats with complex semantics, leading to low testing coverage and reduced fuzzer effectiveness. Although researchers have proposed leveraging Large Language Models (LLMs) to create high-quality seed corpora, current approaches are limited to text-based seed files, such as JavaScript code. To address the limitations, we propose Chat4Seed, a novel approach that extends the capabilities of LLMs to produce not only text but also highly-structured binary seed files. Chat4Seed leverages LLMs to extract and interpret the semantic constraints embedded within formatspecific branches of programs. While existing approaches focus solely on generating text-based seeds, Chat4Seed goes further by utilizing LLMs to generate functional library invocation code to produce binary seeds. It also employs binarylevel manipulation to handle unsupported corner cases, achieving robust seed generation for both text and binary formats. Our experiments and evaluations on 12 real-world programs demonstrate that after 48 hours of fuzz testing using AFL++, the seed corpus generated by Chat4Seed achieves an average increase in coverage of 28.78% compared to traditional crawl-based methods and 39.98% compared to generation-based methods. Additionally, it facilitates the discovery of 84.24% and 392.7% more crashes than seed corpus generated by traditional approaches, respectively, underscoring the potential of Chat4Seed to enhance the efficacy of fuzzing.

The highly dynamic Low Earth Orbit (LEO) environment poses significant challenges for efficient and heterogeneous service provision. A promising approach is to use the communication links of LEO satellites as signals of opportunity, where opportunistic offloading techniques appear to improve the overall performance. However, current solutions often overlook the sophisticated predictive capabilities to exploit spatio-temporal correlation across multiple dimensions (e.g., link quality, node capacity, task requirements), failing to evaluate the quality of transient opportunities for offloading decisions. This paper proposes a Spatio-Temporal Transformer-based Opportunistic Task Offloading (STOTO) approach for heterogeneous tasks in dynamic LEO. The core of STOTO is a spatio-temporal Transformer prediction model, designed to achieve superior awareness of evolving service requirements and resource availability. Then, the scheme dynamically evaluates the situation and resource availability, allocating tasks to the optimal nodes in real time. Experimental results demonstrate that our approach significantly outperforms existing methods, achieving 8.4% higher task completion rates on average.

The ordered escape routing of the differential pair represents a critical component in the physical design of the printed circuit board. An improved Monte Carlo Tree Search (MCTS) algorithm and an innovative technique for estimating escape points are proposed to address the problem of satisfying the constraint of ordered escape routing. Compared with the existing method, the proposed method demonstrates a significant improvement in routing path length and CPU time, both for grid pin arrays and staggered pin arrays. Moreover, the proposed approach effectively addresses the blockage-avoided ordered escape routing problem for differential pairs.

Logs play a crucial role in monitoring system behaviors, providing observability of system events at various levels. However, large-scale software systems generate massive amounts of logs daily, leading to tremendous storage overhead and management burdens. Compression techniques are employed to reduce the storage pressure of logs by eliminating redundancy. Traditional compression methods achieve moderate compression ratios but fail to fully exploit the structural attributes of log data. Several efforts have been made to compress logs based on their structural or temporal attributes, but they struggle with identifying word-level patterns in log contents. Addition-ally, numerical values, such as timestamps, in log headers and parameters pose challenges in compression. To address these limitations, we propose LogCompact, a novel log compressor that recognizes redundancy at both the word and numerical levels. Meanwhile, LogCompact incorporates a word encoder based on Huffman trees to reduce the encoding length of high-frequency words and a two-dimensional differential encoder to simplify the representation of numerical values. Experimental results on five datasets demonstrate the effectiveness of LogCompact in achieving higher compression rates compared to baseline methods.

With the burgeoning prevalence of diabetes and the urgent imperative for effective dietary management, harnessing the power of image analysis and computer vision technology presents a promising solution to the automatic nutrition estimation challenge. However, constructing a comprehensive frame-work for this endeavor presents its own set of challenges, encompassing multiple intricate procedures, including data collection, image segmentation, food recognition, and volume estimation. In this paper, we introduce a sophisticated and automated framework for Chinese food nutrition estimation, leveraging food images sourced from diabetic patients. In addition to developing models for the various procedures, the framework necessitates label images for model training. These label images encompass ground truth masks delineating precise object boundaries for image segmentation, food type labels for food recognition, and comprehensive nutrition tables for each food item. Through rigorous experimentation and validation, our framework has demonstrated its efficacy, offering a convenient and practical tool for managing dietary requirements in diabetes care.

PCB routing becomes time-consuming as the complexity of PCB design increases. Unlike traditional schemes that treat the two essential PCB routing processes separately, namely, escape and bus routing, we consider the continuity between them and present a golden-pin-based routing scheme to find the desired solution with angle and topology constraints. Further, conventional rip-up and reroute methods are often ineffective and inefficient for congestion alleviation and routability optimization. We construct a component graph by modeling components as vertices and applying the minimum weight vertex covering method to improve the routability. A self-adaptable ordering method is presented for escape routing to arrange the pin order on the component boundary, guaranteeing successful bus routing. In addition, escape routing is performed based on a disjoint path method. We construct a dynamic Hanan grid in bus routing and utilize a novel congestion adjustment technique to improve solution quality. Compared with FreeRouting and Allegro, the experiment results show that our algorithm achieves high routability and a significant 90% runtime reduction.

A parallel simulation that partitions a large circuit into sub-circuits is widely used to reduce simulation runtime. To achieve higher simulation throughput, we shall consider signal directions, and thus the final partitioning solution must be acyclic. In this paper, we model a circuit as a directed graph and consider acyclic graph partitioning to minimize edge cuts. This problem differs from the traditional partitioning problem because of the additional acyclicity constraint. Unlike traditional heuristics that tend to be trapped in local minima, especially for large graphs, we present a novel discrete dynamic filled function algorithm for the acyclic graph partitioning problem. Our algorithm can guarantee convergence and effectively move from one discrete local minimizer to another better one. Experimental results show that our algorithm achieves 8% average cutsize reduction over the state-of-the-art works in a comparable runtime.

Efficient exploration in deep cooperative multi-agent reinforcement learning (MARL) still remains challenging in complex coordination problems. In this paper, we introduce a novel Episodic Multi-agent reinforcement learning with Curiosity-driven exploration, called EMC. We leverage an insight of popular factorized MARL algorithms that the ``induced" individual Q-values, i.e., the individual utility functions used for local execution, are the embeddings of local action-observation histories, and can capture the interaction between agents due to reward backpropagation during centralized training. Therefore, we use prediction errors of individual Q-values as intrinsic rewards for coordinated exploration and utilize episodic memory to exploit explored informative experience to boost policy training. As the dynamics of an agent's individual Q-value function captures the novelty of states and the influence from other agents, our intrinsic reward can induce coordinated exploration to new or promising states. We illustrate the advantages of our method by didactic examples, and demonstrate its significant outperformance over state-of-the-art MARL baselines on challenging tasks in the StarCraft II micromanagement benchmark.