AI News

Evaluating the accuracy of KV cache reuse techniques (arxiv.org)

Input-Layer Starvation: Why Per-Layer Pruning Breaks IoT Intrusion Detectors (arxiv.org)

Beyond Mean Attention: Diversity-Aware, Layer-Wise Scoring for KV Cache Eviction (arxiv.org)

Skill Profiling with Attributable Reasoning (SPAR): A Wearable Analysis System for Boxing (arxiv.org)

Query-Conditioned Prototype Adaptation for Cross-Domain Few-Shot Learning: Single-Query Inference, Controlled Comparisons, and Failure Modes (arxiv.org)

Deep-Learning Solvers and Surrogates for Infinity and p-Laplace Problems (arxiv.org)

The KV Cache Is the New Memory Wall (arxiv.org)

NAC: Neural Action Codec for Vision-Language-Action Models (arxiv.org)

Do Location Encoders Capture Spatial Effects? A GeoShapley Benchmark Across Scales (arxiv.org)

Neural Parameter Estimation of RC Thermal Building Models for Model Predictive Control (arxiv.org)

Distributed Learning as a Service: The Developer's Perspective (arxiv.org)

What Do They Fix? LLM-Aided Categorization of Security Patches for Critical Memory Bugs (arxiv.org)

The Communication Map of a Transformer (arxiv.org)

AC Power Flow Contingency Analysis Using a Single Deep Neural Network (arxiv.org)

How Long Does Infinite Width Last? Signal Propagation in Long-Range Linear Recurrences (arxiv.org)

Pseudo-Invertible Neural Networks (arxiv.org)

MeshGraphNet-Transformer: Scalable Mesh-based Learned Simulation for Solid Mechanics (arxiv.org)

FLAME: Flow Enhanced Legendre Memory Models for General Time Series Forecasting (arxiv.org)

Stochastic Bilevel Optimization with Heavy-Tailed Noise (arxiv.org)

Differentially-Private Decision Trees and Provable Robustness to Data Poisoning (arxiv.org)

Tight Stochastic Condition-Number Dependence in Nonconvex-Strongly-Concave Minimax Optimization (arxiv.org)

Retraction-Based Gradient Projection Algorithms on Manifolds (arxiv.org)

Gap-free Differentially Private PCA for Gaussian Data (arxiv.org)

EAServe: Encode-Aware Disaggregated Serving for Multimodal Large Language Models (arxiv.org)

Retrainable physics-integrated neural differentiable modeling of sintering across material systems (arxiv.org)

Retail Product Search: A Practical Approach at Target (arxiv.org)

Aurora-X: Built for Extreme Time Series Forecasting (arxiv.org)

FRESHLATENT: Channel-Aware Latent Adaptation for Resource-Constrained Embodied VLM Perception (arxiv.org)

DiffusionShadow: Diffusion-based Shadow Caching for Neural Volume Rendering (arxiv.org)

StarWM: Self-Supervised Trained Attention Routing for Robust World Models (arxiv.org)

A Comprehensive Study of Content Representations for Speech Synthesis (arxiv.org)

On the Limits of Univariate Deep Learning for Significant Wave Height Forecasting (arxiv.org)

Seeing Speech: Learning Visible Articulatory Dynamics for Speech-Driven 3D Facial Animation (arxiv.org)

AcoustiClaim: A Numeric Claim Benchmark with Instrument Ground Truth (arxiv.org)

BreathGRU: A Novel Semi-Supervised Bidirectional Gated Recurrent Unit Framework for Speech and Breath Segmentation for Respiratory Audio (arxiv.org)

Low-Bit Recurrent States in Hybrid Language Models (arxiv.org)

Towards Understanding Momentum Acceleration in River-Valley Loss Landscape (arxiv.org)

Gradient Surgery for Physics-Informed Neural Networks (arxiv.org)

LipSSM: Structurally Lipschitz-Bounded Cascaded State-Space Model via Metric Transfer between Consecutive SSM Layers (arxiv.org)

Robust Successor Features (arxiv.org)

Metacognitive Selective Ensemble for Mobile Systems (arxiv.org)

Robust Graph Clustering Network for Multiple Missing Data (arxiv.org)

The Residual Stream's Effective Depth (arxiv.org)

Encryptability As a Coordinate Choice: Depth-One Homomorphic Federated Learning of Quantum Neural Networks (arxiv.org)

Scaling Density Functional Theory with Gaussian Splatting (arxiv.org)

Unknown-Traffic Detection, Calibration and Shortcut Reliance in Distilled Encrypted-Traffic Classifiers over One Year (arxiv.org)

Synth-JEPA: Joint Embedding Prediction for Renderer-Free Synthesizer Parameter Search (arxiv.org)

Scaffold-Constrained Subset Dynamic Programming for Exact SSE Clustering (arxiv.org)

Quantizing Looped Transformers: Feedback Exposure and Calibration Blindness (arxiv.org)

LandscapeSHAP: Which Persistent Homology Class Gets the Credit? (arxiv.org)