--Federated Learning (FL) enables collaborative model training across distributed medical institutions while preserving patient privacy, but remains vulnerable to Byzantine attacks and statistical heterogeneity. T o address convergence challenges under data heterogeneity, we develop FedBN-Prox (FedBN-P), combining Federated Batch Normalization with proximal regularization for optimal accuracy-convergence trade-offs. Extensive evaluation across MNIST, CIF AR-10, and Alzheimer's MRI datasets under various Byzantine attack scenarios demonstrates significant improvements over state-of-the-art defenses, achieving up to +1.6 percentage points over FLGuard under non-IID conditions while maintaining robust performance against diverse attack patterns through our adaptive learning approach. In recent years, Federated Learning (FL) has emerged as a powerful paradigm for training deep neural networks across geographically distributed hospitals while preserving patient privacy under stringent regulations such as HIP AA and GDPR [1]-[4]. Recent advances in federated learning for healthcare have shown significant promise in addressing privacy-sensitive medical data challenges through innovative approaches such as secure multi-party computation [5] and blockchain-enhanced frameworks [6] while enabling secure collaborative learning across medical institutions. As illustrated in Figure 1, this collaborative framework allows medical institutions to exchange model updates rather than raw MRI scans, enabling multi-institutional collaboration--for instance, a small rural hospital with just a handful of Alzheimer's MRI scans can still contribute to, and benefit from, a model jointly trained with top-tier research centers. However, real-world FL deployments must cope with two intertwined challenges. First, Byzantine updates --malicious or low-quality gradient submissions--can severely skew the global model and compromise clinical reliability [7], [8], arising from hospitals with insufficient labeled data, poor-quality imaging equipment, or adversarial behavior.

Since its introduction, softmax attention has become the backbone of modern transformer architectures due to its expressiveness and scalability across a wide range of tasks. However, the main drawback of softmax attention is the quadratic memory requirement and computational complexity with respect to the sequence length. By replacing the softmax nonlinearity, linear attention and similar methods have been introduced to avoid the quadratic bottleneck of softmax attention. Despite these linear forms of attention being derived from the original softmax formulation, they typically lag in terms of downstream accuracy. While strong intuition of the softmax nonlinearity on the query and key inner product suggests that it has desirable properties compared to other nonlinearities, the question of why this discrepancy exists still remains unanswered. This work demonstrates that linear attention is an approximation of softmax attention by deriving the recurrent form of softmax attention. Using this form, each part of softmax attention can be described in the language of recurrent neural networks (RNNs). Describing softmax attention as an RNN allows for the ablation of the components of softmax attention to understand the importance of each part and how they interact. In this way, our work helps explain why softmax attention is more expressive than its counterparts.

-- Accurate mass estimation of table-top grown strawberries under field conditions remains challenging due to frequent occlusions and pose variations. This study proposes a vision-based pipeline integrating RGB-D sensing and deep learning to enable non-destructive, real-time and online mass estimation. The method employed YOLOv8-Seg for instance segmentation, Cycle-consistent generative adversarial network (CycleGAN) for occluded region completion, and tilt-angle correction to refine frontal projection area calculations. A polynomial regression model then mapped the geometric features to mass. Experiments demonstrated mean mass estimation errors of 8.11% for not-occluded strawberries and 10.47% for occluded cases. CycleGAN outperformed large mask inpaint-ing (LaMa) model in occlusion recovery, achieving superior pixel area ratios (PAR) (mean: 0.978 vs. 1.112) and higher intersection over union (IoU) scores (92.3% vs. 47.7% in the [0.9-1] range). This approach addresses critical limitations of traditional methods, offering a robust solution for automated harvesting and yield monitoring with complex occlusion patterns. I. INTRODUCTION Fruit mass estimation is essential for optimizing harvest timing, improving agricultural efficiency, and advancing smart, precision agriculture [1].

In this paper, we propose a system for detecting adulteration in coconut milk, utilizing infrared spectroscopy. The machine learning-based proposed system comprises three phases: preprocessing, feature extraction, and classification. The first phase involves removing irrelevant data from coconut milk spectral signals. In the second phase, we employ the Linear Discriminant Analysis (LDA) algorithm for extracting the most discriminating features. In the third phase, we use the K-Nearest Neighbor (KNN) model to classify coconut milk samples into authentic or adulterated. We evaluate the performance of the proposed system using a public dataset comprising Fourier Transform Infrared (FTIR) spectral information of pure and contaminated coconut milk samples. Findings show that the proposed method successfully detects adulteration with a cross-validation accuracy of 93.33%.

Scientific research increasingly depends on complex computational analysis, yet the design and execution of such analysis remain labor-intensive, error-prone, and difficult to scale. From genomics [23, 103, 125], materials discovery [25], drug development [49, 106], and epidemiology [69, 9], to climate modeling [60] and personalized medicine [36], the analytical backbone of modern science involves highly structured, multi-step workflows written in code. These workflows must integrate raw or semi-structured data, apply statistical or mechanistic models, and yield interpretable results--all under the constraints of evolving domain knowledge, platform variation, and methodological rigor. Recent advances in large language models (LLMs) have led to rapid progress in general-purpose agents capable of decomposing tasks [118, 113, 79, 18], interacting with tools [86, 138, 147, 49], and producing executable code [30, 126, 167, 114, 140]. These developments have raised the prospect that LLM-based agents might soon contribute meaningfully to scientific discovery by automating analysis pipelines, exploring hypotheses, or refining computational models [8, 60]. However, realizing this promise requires bridging a fundamental gap between general reasoning ability and the structured, precision-driven nature of scientific computation. While many LLM-based agents have shown competence in retrieving documents, calling APIs, or planning abstract tasks [165, 154, 46, 124], these capabilities fall short in domains where scientific progress depends on code. In fields such as transcriptomics [90, 69, 9], protein engineering [106], and statistical genetics [36, 76], research workflows are encoded as sequences of programmatic transformations, each tailored to the idiosyncrasies of a specific dataset, model assumption, or experimental design.

Attributed graph clustering holds significant importance in modern data analysis. However, due to the complexity of graph data and the heterogeneity of node attributes, leveraging graph information for clustering remains challenging. To address this, we propose a novel deep graph clustering model, GCL-GCN, specifically designed to address the limitations of existing models in capturing local dependencies and complex structures when dealing with sparse and heterogeneous graph data. GCL-GCN introduces an innovative Graphormer module that combines centrality encoding and spatial relationships, effectively capturing both global and local information between nodes, thereby enhancing the quality of node representations. Additionally, we propose a novel contrastive learning module that significantly enhances the discriminative power of feature representations. In the pre-training phase, this module increases feature distinction through contrastive learning on the original feature matrix, ensuring more identifiable initial representations for subsequent graph convolution and clustering tasks. Extensive experimental results on six datasets demonstrate that GCL-GCN outperforms 14 advanced methods in terms of clustering quality and robustness. Specifically, on the Cora dataset, it improves ACC, NMI, and ARI by 4.94%, 13.01%, and 10.97%, respectively, compared to the primary comparison method MBN.

This paper aims to develop a Machin e Learning (ML) - based system for detecting honey adulteration utilizing honey mineral element profiles. The proposed system comprises two phases: preprocessing and classification. The preprocessing phase involves the treatment of missing - value attributes a nd normalization. In the classification phase, we use three supervised ML models: logistic regression, d ecision tree, and random forest, to discriminate between authentic and adulterated honey. To evaluate the performance of the ML models, we use a public dataset comprising measurements of mineral element content of authentic honey, sugar syrups, and adulterated honey. Experimental findings show that mineral element content in honey provides robust discriminative information for detecting honey adulteration . Results also dem onstrate that the random forest - based classifier outperforms other classifiers on this dataset, achieving the highest cross - validation accuracy of 98.37%.

While 3D Gaussian Splatting (3DGS) has revolutionized photorealistic mapping, conventional approaches based on camera sensor, even RGB-D, suffer from fundamental limitations such as high computational load, failure in environments with poor texture or illumination, and short operational ranges. LiDAR emerges as a robust alternative, but its integration with 3DGS introduces new challenges, such as the need for exceptional global alignment for photorealistic quality and prolonged optimization times caused by sparse data. To address these challenges, we propose GSFusion, an online LiDAR-Inertial-Visual mapping system that ensures high-precision map consistency through a surfel-to-surfel constraint in the global pose-graph optimization. To handle sparse data, our system employs a pixel-aware Gaussian initialization strategy for efficient representation and a bounded sigmoid constraint to prevent uncontrolled Gaussian growth. Experiments on public and our datasets demonstrate our system outperforms existing 3DGS SLAM systems in terms of rendering quality and map-building efficiency.

Due to growing privacy concerns, machine unlearning, which aims at enabling machine learning models to ``forget" specific training data, has received increasing attention. Among existing methods, influence-based unlearning has emerged as a prominent approach due to its ability to estimate the impact of individual training samples on model parameters without retraining. However, this approach suffers from prohibitive computational overhead arising from the necessity to compute the Hessian matrix and its inverse across all training samples and parameters, rendering it impractical for large-scale models and scenarios involving frequent data deletion requests. This highlights the difficulty of forgetting. Inspired by cognitive science, which suggests that memorizing is easier than forgetting, this paper establishes a theoretical link between memorizing (incremental learning) and forgetting (unlearning). This connection allows machine unlearning to be addressed from the perspective of incremental learning. Unlike the time-consuming Hessian computations in unlearning (forgetting), incremental learning (memorizing) typically relies on more efficient gradient optimization, which supports the aforementioned cognitive theory. Based on this connection, we introduce the Influence Approximation Unlearning (IAU) algorithm for efficient machine unlearning from the incremental perspective. Extensive empirical evaluations demonstrate that IAU achieves a superior balance among removal guarantee, unlearning efficiency, and comparable model utility, while outperforming state-of-the-art methods across diverse datasets and model architectures. Our code is available at https://github.com/Lolo1222/IAU.

Contextual hallucinations -- statements unsupported by given context -- remain a significant challenge in AI. We demonstrate a practical interpretability insight: a generator-agnostic observer model detects hallucinations via a single forward pass and a linear probe on its residual stream. This probe isolates a single, transferable linear direction separating hallucinated from faithful text, outperforming baselines by 5-27 points and showing robust mid-layer performance across Gemma-2 models (2B to 27B). Gradient-times-activation localises this signal to sparse, late-layer MLP activity. Critically, manipulating this direction causally steers generator hallucination rates, proving its actionability. Our results offer novel evidence of internal, low-dimensional hallucination tracking linked to specific MLP sub-circuits, exploitable for detection and mitigation. We release the 2000-example ContraTales benchmark for realistic assessment of such solutions.