7103 papers
This paper introduces FedSCS-XGB, a novel federated learning protocol for distributed XGBoost that preserves the algorithm's structural advantages and achieves near-centralized performance (within a 1% gap) for continual health monitoring via wearable sensors.
DC-Merge is a novel model merging method that achieves state-of-the-art performance by first balancing the energy distribution of task vectors through singular value smoothing and then aligning their directional geometries via projection onto a shared orthogonal subspace to preserve multi-task knowledge.
This paper analyzes the gradient flow dynamics of the value-softmax model, a core component of self-attention, to demonstrate that training inherently drives the system toward low-entropy solutions, thereby providing a theoretical explanation for empirical phenomena like attention sinks and massive activations.
This paper proposes SPPCSO, an adaptive penalized estimation method that integrates single-parametric principal component regression with regularization to achieve stable variable selection and robust coefficient estimation in high-dimensional, highly correlated, and noisy datasets, outperforming traditional approaches in both theoretical bounds and practical applications such as gene discovery.
This paper introduces Synthetic Monitoring Environments (SMEs), a configurable suite of continuous control tasks with known optimal policies and exact regret metrics, designed to enable rigorous, white-box diagnostics and systematic analysis of reinforcement learning algorithms' performance under varying conditions.
This paper proposes the BAEN-SVM, a robust binary classification model utilizing a novel bounded asymmetric elastic net loss function that effectively handles noise and geometric irregularities through theoretical guarantees of bounded influence and Fisher consistency, solved via an efficient half-quadratic algorithm.
This paper proposes DeCoST, a learning-based two-stage framework that effectively decouples discrete and continuous variables to solve the Orienteering Problem with Time Windows and Variable Profits, achieving superior solution quality and significant inference speedups compared to state-of-the-art methods.
This study demonstrates that agentic retrieval-augmented reasoning pipelines significantly enhance the collective reliability, consensus strength, and cross-model robustness of large language models in radiology question answering compared to zero-shot inference, while highlighting that accuracy and agreement alone are insufficient metrics for evaluating clinical safety under model variability.
This paper presents a neural network implementation of Fuzzy Cognitive Maps (FHM) that utilizes Langevin differential dynamics to learn causality patterns, derive inverse solutions for output modification, and evaluates its performance across multiple datasets.
This paper introduces Stem, a novel plug-and-play sparse attention module that overcomes the quadratic complexity bottleneck in long-context LLMs by aligning sparsity with causal information flow through position-dependent token retention and an output-aware metric, thereby achieving superior accuracy with reduced computational cost and latency.
The paper introduces GMM-PIELM, a probabilistic adaptive sampling framework that significantly improves the accuracy and conditioning of Physics-Informed Extreme Learning Machines for stiff PDEs by autonomously concentrating basis function centers in high-error regions like shock fronts, achieving orders-of-magnitude lower errors than baseline methods while retaining rapid closed-form training speeds.
This paper proposes a novel 3D dental implant inpainting method using perpendicular score-based diffusion models that operate in the projection domain to capture inter-projection correlations, thereby generating high-quality, artifact-reduced CBCT images with improved consistency compared to existing 2D-based approaches.
This paper introduces Polarized Direct Cross-Attention (PolaDCA), a novel relational learning framework that overcomes the limitations of static graph structures in conventional GNNs by employing data-driven, adaptive message passing to achieve state-of-the-art fault diagnosis accuracy and noise robustness in rotating machinery.
This paper proposes a three-stage post-training pipeline that computes fine-grained entropy-based uncertainty, calibrates them via Platt scaling, and uses reinforcement learning to teach language models to efficiently generate interpretable and well-calibrated uncertainty estimates at test time, outperforming existing post-hoc methods in both performance and generalization.
The paper introduces SAHOO, a practical framework that employs a Goal Drift Index, constraint preservation checks, and regression-risk quantification to effectively monitor and control alignment drift while significantly improving performance in recursive self-improving systems across code, reasoning, and truthfulness tasks.
The paper introduces AIRT, an end-to-end deep-learning framework that generates high-quality, deliverable single-arc VMAT prostate treatment plans in under one second directly from CT images and contours, demonstrating non-inferiority to standard clinical planning systems while significantly accelerating workflow efficiency.
MoEless is a novel serverless serving framework that mitigates expert load imbalance in Mixture-of-Experts (MoE) large language models through proactive load prediction and dynamic scaling, achieving significant reductions in inference latency and cost compared to existing solutions.
The paper introduces the Dynamic Chunking Diffusion Transformer (DC-DiT), a model that adaptively compresses image tokens based on spatial detail and diffusion timesteps to achieve superior generation quality and efficiency over standard DiTs while requiring minimal additional training.
The paper introduces the Frequency-Separable Hamiltonian Neural Network (FS-HNN), a novel architecture that decomposes Hamiltonian functions into distinct fast and slow modes trained on different timescales to overcome the spectral bias of existing methods and significantly improve long-horizon extrapolation for multi-timescale dynamical systems and PDEs.
This paper proposes a hardware-independent, client-side classification framework that leverages the Normalized Compression Distance (NCD) within kernel methods to achieve high accuracy on minimal data without requiring large centralized datasets or extensive computational resources.