AI News

Mitigating Reward Hacking in Realistic Scenarios Through Representation Engineering (arxiv.org)

MHeatmap: Two-Walk Laplacian Reordering and Mosaic Layout for Rectangular Matrix Visualization (arxiv.org)

Causal world models under physical parameter shifts (arxiv.org)

Value-Rectified Distillation for Flow-based Offline Reinforcement Learning (arxiv.org)

The Price of Choice in Selective Prediction (arxiv.org)

Distilling Multi-Teacher Scoring Principles for Unsupervised Time Series Anomaly Detection (arxiv.org)

Dynamics of Hidden Learning Structure (arxiv.org)

Covariate-assisted graph matching (arxiv.org)

ART-MPG: Autoregressive R/T Manifold Path Guiding (arxiv.org)

Gate Initialisation versus Recurrent Dropout and Zoneout in Learning Long Dependencies: A Controlled Study of LSTM and GRU Networks on Synthetic Tasks (arxiv.org)

Better Policies, Worse Verifier Audits (arxiv.org)

Dissimilarity-Based Markov Decision Processes for Stabilizing State Distributions (arxiv.org)

Local and Sparse Kernel Operator Learning for Incompressible Flows: An Operator Lagrange Formulation (arxiv.org)

FlashNext-RL: A System Optimization Study for Long-Horizon Agentic RL of Qwen3.8-Flash-Next (arxiv.org)

Rethinking Time Series Tokenization from a Frequency Perspective (arxiv.org)

Doob-Based Generative Policy Optimization (arxiv.org)

Attention-based Classification for Indirect Prompt Injection (arxiv.org)

SSMCGM-Stream: A Streaming State-Space Model for Personalized Diabetes Management (arxiv.org)

How Does Orthogonalization Adapt to the Neural Network Hessian Structure? A Gradient Self Outer-Product Analysis at Initialization (arxiv.org)

From Weather to Enhanced Weathering: Weather Foundation Models for Carbon Dioxide Removal at Operational Scale (arxiv.org)

Unit-Aware Differentially Private Quantum Machine Learning for Clinical and Wearable Health Data (arxiv.org)

Transformers as Generic Deep Neural Networks: Exact Compilation, the Cost of Removing Activation-Activation Products, and Multi-Agent Attention Clustering (arxiv.org)

Is 2:4 Sparsity Still Worth It? Semi-Structured Sparsity versus Quantization in LLM Inference (arxiv.org)

Multi-Objective Protein Optimization via Modular Diffusion Steering (arxiv.org)

The Data Differentiator Framework (DDF): Quantifying Marginal Data Value with Information-Theoretic and Game-Theoretic Measures in an Agentic Evaluation Pipeline (SDSAE) (arxiv.org)

Query Feedback and Teaching for Stable and Certifiable Language Generation (arxiv.org)

A Model-Free Continuous-Time RL Algorithm from Discrete-Time Off-Policy Data with End-to-End Guarantees (arxiv.org)

Minimax-Optimal Predictive Flow Matching under Distribution Drift (arxiv.org)

Combating Catastrophic Forgetting in Continual Domain Adaptation via Knowledge Recasting (arxiv.org)

RL-IF: Tracking Non-local Training Data Influence for Online Reinforcement Learning (arxiv.org)

LEAN: Loss-Engineered Modulation of the Model Weight-Sensitivity Distribution for Efficient, Error-Tolerant Neural Networks (arxiv.org)

How Much Gradient Does Truncated Backprop Lose in Looped Transformers? (arxiv.org)

A General Topological Symmetry Index for Critical Orbits In Neural Network Loss Landscapes (arxiv.org)

When Do Gaussian Priors Fail? A Latent Geometry Investigation of RL Fine-Tuning (arxiv.org)

A Mechanistic Study of Automated Failure Attribution in LLM-based Agentic Systems (arxiv.org)

Towards Realistic Models of Change: A Causal Perspective on Decision-Making and Dynamical Systems (arxiv.org)

An Explicit Feedback Contract for Iterative ML-Agent Harnesses (arxiv.org)

CALLIOPE: Private Insights into AI Use, Without Embeddings (arxiv.org)

ArcFlow: Training Flow Policies for Multiple Solver Budgets (arxiv.org)

Measure First, Then Allocate: Calibrated Rollout Budgets for MCP Agents in Reinforcement Learning (arxiv.org)

Exploiting Concavity Information in Gaussian Process Contextual Bandit Optimization (arxiv.org)

Geometric Algebra for Neural Networks: A Survey (arxiv.org)

Is Refusal Direction Universal across Low-Resourced Languages? (arxiv.org)

The role of locality and weight sharing in convolutional neural networks (arxiv.org)

Mitigating Model Collapse through Dreaming Learning (arxiv.org)

Mastering Model Fit: Applications of the InterModel Vigorish for Interpretable Machine Learning (arxiv.org)

How Do DiTs Denoise? An Aggregator-Predictor View of Attention and MLPs (arxiv.org)

Asymmetric Likelihood-Free Hypothesis Testing (arxiv.org)

Post-hoc Epistemic Uncertainty for Vision-Language-Action Policies (arxiv.org)

Fragment-Level Graph Encoding for Interpretable Molecular Property Prediction (arxiv.org)