AI News

Bifurcations from Weight Space: JEPAWG on Parametric Dynamical Systems (arxiv.org)

Reverse-KL Policy Distillation: Statistical Guarantees and Limits (arxiv.org)

Lean Finder V2: Semantic Search for Mathlib That Disambiguates User Queries (arxiv.org)

Neural Spectral Capacity with Residual Context for Heterogeneous Resource Allocation within Transformers (arxiv.org)

Is spectral predictability enough? Testing {\Omega}-matching for synthetic pretraining under distribution shift (arxiv.org)

A Computationally Efficient and Rate-Optimal Algorithm for Adversarial Linear MDPs with (Aggregate) Bandit Feedback (arxiv.org)

Escaping the Phylogenetic Trap: Persistent Homology Enables Cross-Species Ion Channel Classification via Multimodal Graph Neural Networks (arxiv.org)

Agent AI Zarz\k{a}dzaj\k{a}cy Sieci\k{a} Open RAN Zorientowana na U\.zytkownika (arxiv.org)

How to Use Parallel Data for Multilingual Pretraining (arxiv.org)

CompLit: Scientific Literature Search Benchmarks Must Cover Implicit, Cumulative, and Unmet Needs (arxiv.org)

Context-Augmented, Session-Calibrated Blood Pressure Tracking from Photoplethysmography (arxiv.org)

Same Answer, Different Reasons: Trace-Level Divergence in Multilingual Large Language Models (arxiv.org)

Bridging the Data Gap: Source-Informed Task-Space Transport for Few-Shot Diffusion (arxiv.org)

Harmful Responses Are Jailbreak Templates: Transferring Autoregressive Jailbreaks to Diffusion Language Models (arxiv.org)

SQuAT: Spherical Lattice Vector Quantization-Aware Training (arxiv.org)

PYTHIA, a transferable machine learning framework for simultaneous defect and adsorption site optimization (arxiv.org)

OP-Print: Tracing Membership Imprints in On-Policy Self-Distillation (arxiv.org)

GRAM: Group-wise Rank-Aware Model Merging via Subspace Alignment (arxiv.org)

StochIKM: Automated High-Fidelity Reconstruction of Individual Patient Data from Published Kaplan-Meier Curves (arxiv.org)

NormTune: Norm-Guided Parameter-Efficient Fine-Tuning for Reasoning-Intensive Tasks (arxiv.org)

LaVGuard: Selective Decoding for Contextual Value Drift (arxiv.org)

Beyond source attribution: Address authorship and comparability decisions in an LLM-built scholarly knowledge graph (arxiv.org)

Charts as Epistemic Actions for Experience-Driven Agentic Data Analysis (arxiv.org)

Same Luck, Different Choice: Twin-World Credit Assignment for Agentic Reinforcement Learning (arxiv.org)

Shift-aware score transport for privacy auditing without retraining (arxiv.org)

Fine-Grained Analysis of Catastrophic Forgetting in Continual Learning via Algorithmic Stability (arxiv.org)

Efficient and Optimal Learning of Discrete Distributions with Person-Level Privacy (arxiv.org)

Questioning Instrumental Convergence in RLVR: Checkpoint Evidence from OLMo-3 (arxiv.org)

Graph-Recurrent Learning of Emergent Collective Dynamics (arxiv.org)

Learning Checklists to Reduce Lottery Effects in Scientific Peer Review (arxiv.org)

Who Verifies the Builder? Toward Independent Assurance for Agent-Produced Software (arxiv.org)

VHDLBench: Benchmarking and Fine-Tuning Large Language Models for VHDL-Based Chip Design (arxiv.org)

VerifierBench: Benchmarking the Generation of Executable Verifiers for LLM Agents (arxiv.org)

Where Should a Stuck Agent Spend Its Next Tokens? Causal Attribution of Inference-Time Resources (arxiv.org)

BeNice! Closing Looping's Inference-Memory Gap via KV-Sharing (arxiv.org)

Causal Chain of Thought: Mitigating Attention Leakage via Context Isolation (arxiv.org)

A Fix for Softmax and Cross-Entropy (arxiv.org)

METTA: Memory-Enhanced Test-Time Adaptation for Meta-Reinforcement Learning (arxiv.org)

Cost-Aware Abstention for LLM-Based IT Ticket Classification (arxiv.org)

Do Machine Learning-Guided Proactive Full Capacity Protocols Mitigate Emergency Department Overcrowding? A Simulation-Based Evaluation (arxiv.org)

The Cost of Saying Hello: An Empirical Study of Handshake Overhead in AI Agent Communication Protocols (arxiv.org)

baseVultra: Compact Nonlinear Recurrence with Reallocated Capacity (arxiv.org)

IRAH: Hierarchical Agentic Tree Search for Retrieval Pipeline Optimization (arxiv.org)

Detecting AI Training in Datacentre Power (arxiv.org)

SIGNN++: Robust Graph Neural Star Identification on Earth (arxiv.org)

ReDence: A Hybrid Agentic and Decision-Model Framework for Rapid, Evidence-Grounded Detection of Modern Social Engineering Scams (arxiv.org)

Joint or Per-Option? Compact Encoders for Closed-Set Decisions from 17 Million to 1 Billion Parameters (arxiv.org)

Bridging the Closed-World Gap: Interpretable Anomaly Detection via a Hybrid TSK-FCM Ensemble on the BETH Dataset (arxiv.org)

Spectral Uncertainty Decomposition for Sparse Bayesian Learning (arxiv.org)

Short-to-Long: Detached Features for Long-Horizon Value Learning (arxiv.org)