AI News

Representational Capacity: Geometric Limits on Feature Representation in Transformer Language Models (arxiv.org)

Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models (arxiv.org)

KVarN: Variance-Normalized KV-Cache Quantization Mitigates Error Accumulation in Reasoning Tasks (arxiv.org)

Optimal and exact recovery on the general nonuniform Hypergraph Stochastic Block Model (arxiv.org)

Conformal Language Modeling via Posterior Sampling (arxiv.org)

WISE-HAR: A Generalizable Ensemble Deep Learning Framework for WiFi-Based Human Activity Recognition (arxiv.org)

A Sparse Bayesian Learning Algorithm for Estimation of Interaction Kernels in Motsch-Tadmor Model (arxiv.org)

Inducing Reasoning Primitives from Agent Traces (arxiv.org)

Efficient ASR Training with Conversations that Never Happened (arxiv.org)

Learning Coherent Representations: A Topological Approach to Interpretability (arxiv.org)

CoMPAS3D: A Dataset and Benchmark for Interactive Motion (arxiv.org)

RRISE: Robust Radius Inference via a Surrogate Estimator (arxiv.org)

NetKV: Network-Aware Decode Instance Selection for Disaggregated LLM Inference (arxiv.org)

DELTAMEM: Incremental Experience Memory for LLM Agents via Residual Trees (arxiv.org)

dLLM-Cache: Accelerating Diffusion Large Language Models with Adaptive Caching (arxiv.org)

Neural Posterior Estimation for Stochastic Epidemic Models Using Final Outcome Data (arxiv.org)

ToolGate: Token-Efficient Pre-Call Control for Tool-Augmented Vision-Language Agents (arxiv.org)

The Impact of Configuring Agentic AI Coding Tools on Build-vs-Buy Decisions: A Study Protocol (arxiv.org)

High-Dimensional Latents Should Be Diagnosed Through Phase Structure (arxiv.org)

scTranslation: A Comprehensive Benchmark for Single-Cell Multi-Omics Modality Translation (arxiv.org)

Hedge-Bench: Benchmarking Agents on Hard, Realistic Tasks Pertaining to Financial Reasoning (arxiv.org)

Local and Global Contraction Principles for MCMC Mixing (arxiv.org)

Agent libOS: A Library-OS-Inspired Runtime for Long-Running, Capability-Controlled LLM Agents (arxiv.org)

TRAP: Hijacking VLA CoT-Reasoning via Adversarial Patches (arxiv.org)

ERP-XTTN: Interpretable Prototype-Guided Cross-Attention for Cross-Subject ERP Classification (arxiv.org)

Fast Unlearning at Scale via Margin Self-Correction (arxiv.org)

Hierarchical RBF-KAN and RBF-SKAN Architectures for Multidimensional Function Approximation and Random Field Learning (arxiv.org)

The Reliability Gap in Benchmark Auditing: Distribution Shift and Scale as Failure Modes of Contamination Detection (arxiv.org)

Gate AI: LLM Security Benchmark Evaluation Methodology and Results (arxiv.org)

Multi$^2$: Hierarchical Multi-Agent Decision-Making with LLM-Based Agents in Interactive Environments (arxiv.org)

The Violation Situation Pattern: A Knowledge-Graph Pattern for Compliance Violations (arxiv.org)

CP-Agent: Context-Aware Multimodal Reasoning for Cellular Morphological Profiling under Chemical Perturbations (arxiv.org)

Position: Adversarial ML for LLMs Is Not Making Any Progress (arxiv.org)

SEAOTTER: Sensor Embedded Autoencoding with One-Time Transcode for Efficient Reconstruction (arxiv.org)

CoralBay: A Self-Supervised CT Foundation Model (arxiv.org)

PyraMathBench: Evaluating and Improving Mathematical Capability in Large Language Models (arxiv.org)

Improvise, Adapt, Overcome: An On-The-Fly Multifidelity Algorithm for Efficient Machine Learning (arxiv.org)

Building Trust in Black-box Optimization: A Comprehensive Framework for Explainability (arxiv.org)

Spectral Asymptotics of Neural Network Loss Landscapes: An Exact Decomposition of the Curvature Exponent (arxiv.org)

R2DN: Scalable Parameterization of Contracting and Lipschitz Recurrent Deep Networks (arxiv.org)

Reasoning Structure of Large Language Models (arxiv.org)

MOSAIC: Efficient Mixture-of-Agent Scheduling via Adaptive Aggregation and Inference Concurrency (arxiv.org)

Large Byte Model: Teaching Language Models About Compiled Code (arxiv.org)

AUGUSTE: Online-Learning dApp for Predictive URLLC Scheduling (arxiv.org)

Anomalies in Multivariate Time Series Benchmarks Are Mostly Univariate (arxiv.org)

A formal definition and meta-model for a machine theory of mind (arxiv.org)

Cross-Lingual Token Arbitrage: Optimizing Code Agent Context Windows via Local LLM Preprocessing (arxiv.org)

What Do Students Learn? A Feature-Level Analysis of Dark Knowledge (arxiv.org)

Learning to Solve, Forgetting to Retain: Correct-Set Turnover in RLVR (arxiv.org)

AnchorMoE: Interpretable Time Series Classification via Anchor-Routed MoE (arxiv.org)