AI News

Still: Amortized KV Cache Compaction in a Single Forward Pass (arxiv.org)

Late-Layer Fusion is Enough: Dual-Path Vision Token Routing for Multimodal Large Language Models under Visual Saturation (arxiv.org)

Frequency-based Constrained Sampling for Interval Patterns (arxiv.org)

VESTA: A Fully Automated Scenario Generation and Safety Evaluation Framework for LLM Agents (arxiv.org)

Scaffold Effects on GAIA: A Controlled Comparison (arxiv.org)

Standpoint Logics with Defeasible Beliefs (arxiv.org)

Explaining Black-Box Language Models: Learning to Optimize Linguistically-Structured Word Subsets (arxiv.org)

Prescriptive Scaling Reveals the Evolution of Language Model Capabilities (arxiv.org)

On the Complexity of Offline Reinforcement Learning with $Q^\star$-Approximation and Partial Coverage (arxiv.org)

More Bang for the Buck: Improving the Inference of Large Language Models at a Fixed Budget using Reset and Discard (ReD) (arxiv.org)

Self-Supervised Dynamical System Representations for Physiological Time-Series (arxiv.org)

A Robust $\widetilde{\mathcal{O}}(1/\sqrt{T})$ Rate for Unprojected TD Learning with Linear Function Approximation (arxiv.org)

Adaptive Resolution for Finite-Rank Gaussian Processes (arxiv.org)

Joint stochastic localization and applications (arxiv.org)

Dynamics of learning to integrate in linear recurrent neural networks (arxiv.org)

Investigating the Histogram Loss in Regression (arxiv.org)

The Mirrored Influence Hypothesis: Efficient Data Influence Estimation by Harnessing Forward Passes (arxiv.org)

Consensus-based adaptive sampling and approximation for high-dimensional energy landscapes (arxiv.org)

From 0-to-1 to 1-to-N: Reproducible Engineering Evidence for MetaAI Recursive Self-Design (arxiv.org)

Semi-Parametric Inference for Doubly Stochastic Spatial Point Processes: An Approximate Penalized Poisson Likelihood Approach (arxiv.org)

Partial Identification under Missing Data Using Weak Shadow Variables from Pretrained Models (arxiv.org)

Wedge Sampling: Efficient Tensor Completion with Nearly-Linear Sample Complexity (arxiv.org)

Performative Learning Theory (arxiv.org)

Extending Ontologies: From Dense Embeddings to Hybrid Quantum-Fuzzy Systems (arxiv.org)

Improved Analysis of the Accelerated Noisy Power Method with Applications to Decentralized PCA (arxiv.org)

Full-Batch Gradient Descent Outperforms One-Pass SGD: Sample Complexity Separation in Single-Index Learning (arxiv.org)

Conditional Normalizing Flows for Forward and Backward Joint State and Parameter Estimation (arxiv.org)

Locally Adaptive Conformal Inference for Operator Models (arxiv.org)

LARP: Learner-Agnostic Robust Data Prefiltering (arxiv.org)

Robust Random Graph Matching in Dense Graphs via an Approximate Message Passing Type Algorithm (arxiv.org)

Entropic Optimal Transport Eigenmaps for Nonlinear Alignment and Joint Embedding of High-Dimensional Datasets (arxiv.org)

Bandits for Efficient Experimentation: Adapting to Control Group, Preferences, and Context Drifts (arxiv.org)

What Makes a Desired Graph for Relational Deep Learning? (arxiv.org)

Testing the Black Box: Structural Barriers to Independent Evaluation of Consumer-Facing Health LLMs (arxiv.org)

A Variability-Based Framework for Interpretable Naming in Formal and Relational Concept Analysis (arxiv.org)

SpectrumKV: Per-Token Mixed-Precision KV Cache Transfer for Prefill-Decode Disaggregated LLM Serving (arxiv.org)

Tyan-WP: A Wind Power Foundation Model for Ultra-Short-Term Probabilistic Forecasting (arxiv.org)

Reinforcement Learning for Flow-Matching Policies with Density Transport (arxiv.org)

Teacher-Free Self-Training Amplifies but Does Not Compound: A Pass@$K$ Crossover on a Free-Verifier Domain (arxiv.org)

Towards Long-Horizon Vessel Trajectory and Destination Forecasting with Reasoning Large Language Models (arxiv.org)

GIFT: LLM-Guided State-Reward Interface for Financial Reinforcement Learning (arxiv.org)

Trajectory-Refined Distillation (arxiv.org)

Self-Evolving Scientific Agent Discovers Generalizable Physically-Reasoned Fluid Control (arxiv.org)

TT-DAC-PS: Twin-Target Deterministic Actor-Critic with Policy Smoothing for Optimal Trade Execution (arxiv.org)

On the Wasserstein Geodesic Principal Component Analysis of probability measures (arxiv.org)

Byzantine Cheap Talk: Adversarial Resilience and Topology Effects in LLM Coordination Games (arxiv.org)

Contrast encodes inductive bias: separating slow noise from dynamics in predictive representation learning (arxiv.org)

scCBGM: Interpretable Single-Cell Counterfactual Editing (arxiv.org)

Characterizing the Discrete Geometry of ReLU Networks (arxiv.org)

GenTSE: Enhancing Target Speaker Extraction via a Coarse-to-Fine Generative Language Model (arxiv.org)