AI News

ROUTEAUDIT: Interaction-Aware Identification for Budgeted Multi-Verifier Routing (arxiv.org)

iS-KV: Online Low-Rank KV Cache Compression via Block-Incremental SVD (arxiv.org)

MetaRubric: Learning to Reward for Rubric-Based Reinforcement Learning (arxiv.org)

Training-Loss Guarantees for Muon with Finite-Step Newton--Schulz Orthogonalization (arxiv.org)

Cordial Learning: Distributed Training with Correlated Data (arxiv.org)

Follow the Winners: Conservative Policy Improvement with the Cross-Entropy Method for Critic-Free RFT (arxiv.org)

Operator-informed initialization for Fourier features physics-informed neural networks (arxiv.org)

LUMOS: Tracing Parametric Knowledge from Training Data to Behavioral Outputs in LLMs (arxiv.org)

Scaling Trajectories for Complex Tasks through Recursive Self-Rewrite (arxiv.org)

MLCommons Jailbreak Benchmark v1.0 (arxiv.org)

FSPO: Policy-Consistent Risk and Pareto-Feasible Control for Budgeted LLM RL Post-Training (arxiv.org)

AMBER: Multi-View Adaptive Budget Allocation for Listwise Vision-Language Reranking (arxiv.org)

DNAlign: Dynamic Null-Space Safe Alignment for LLMs (arxiv.org)

Bidirectional Voronoi-biased Exploration Curriculum for Reinforcement Learning (arxiv.org)

A Unified Framework for Bayesian Data Assimilation with Generative Models and Observation Interpolants (arxiv.org)

16-bit Precision of Convolutional Neural Networks on Microcontroller Units for 8-bit Costs (arxiv.org)

AIBL: Augmented Instance-Based Learning with Structured Memory and Neural Embeddings (arxiv.org)

Most-Recent Anchoring with Recurrent Ordering for Time Series Forecasting (arxiv.org)

Rethinking Epistemic Uncertainty in Node Classification through Information Growth (arxiv.org)

Bounded Reachability & Jailbreak Detection via Contraction-Constrained State Space Models (arxiv.org)

Harness-Aware Distillation for Small Language Model Agents (arxiv.org)

TACD: Distilling Efficient Text-to-Motion Models via Terminal Amplification Control (arxiv.org)

PsyEvo: A Personalized Counseling Agent That Self-Evolves at Test Time (arxiv.org)

OptiSelect: How does the Optimizer Shape Data Curriculum? (arxiv.org)

Causal Representation Learning with Instantaneous and Lagged Relations via Nonstationarity (arxiv.org)

Measure Less, Know More: Self-Supervised Test-Time Feature Acquisition (arxiv.org)

Beyond Random Splits: Evaluating Drug-Target Affinity Models Under Chemically and Biologically Motivated Distribution Shifts Copy (arxiv.org)

Continual Graph Memory for Mathematical Research Agents (arxiv.org)

Frequency Is Not Sensitivity Identifying Safety-Sensitive Experts in Sparse MoE LLM (arxiv.org)

HASTE: Evolving Agent Harnesses Against Emerging Attacks Using Sparse Evidence (arxiv.org)

Positive-Unlabeled Learning for Agent Safety False Alarm Auditing (arxiv.org)

When to Compile a Computer-Use Agent? Measuring Payback and Making Compilation Decisions for Token Efficiency (arxiv.org)

Single or Multiple Policies for Phase-Structured Reinforcement Learning? (arxiv.org)

Metropolis-Hastings Dominates Importance Resampling for Policy Composition (arxiv.org)

Dual-Context Analog Retrieval for Time Series Forecasting (arxiv.org)

Certified Mechanistic Edits: Behavioral Guarantees for Skill Removal and Preservation (arxiv.org)

Reasoning with Evidence, Not Merely Rationales: Verifiable Preference Proofs for LLM-Based Recommendation (arxiv.org)

Distributed Learning with Selective State Space Models: Architecture-Aware Convergence Analysis (arxiv.org)

Reliable Self-Evolution with Imperfect Proxy Rewards (arxiv.org)

Relevant Evidence Decoding for Audio-Visual Hallucination Mitigation (arxiv.org)

Getting Your Guidance Weights Right in diffusion and flow-matching posterior sampling (arxiv.org)

An Automated and Reproducible Workflow for Crack Identification and Damage Assessment of Fusion Materials (arxiv.org)

Beyond Trained Models: Compiling GNNs for a Sound Explainer Benchmark (arxiv.org)

Divergence controls entropy in distillation (arxiv.org)

RASPER: Reward-Aligned Summarization of Clinical Notes for EHR Outcome Prediction (arxiv.org)

Safeguarding Mutual Correction in Source-Free Domain Adaptation via Cut Statistics (arxiv.org)

PLCWorld: Benchmarking LLM-Generated PLC Programs in Closed-Loop Plant Simulation (arxiv.org)

Personalized Automatic Speech Recognition for a Dysarthric and Tracheostomic Speaker using Artificial Conversations (arxiv.org)

ZeroMAG: Zero-Shot Multimodal Adapter Generation for Plug-and-Play EEG Foundation Models (arxiv.org)

Objects Without Morphisms: What LLMs for Mathematics Do Not Represent (arxiv.org)