AI News

OPTIMA: Optimal One-shot Pruning for LLMs via Quadratic Programming Reconstruction (arxiv.org)

Reinforced Linear Genetic Programming (arxiv.org)

Activation Sensitivity as a Unifying Principle for Post-Training Quantization (arxiv.org)

Expected Return Causes Outcome-Level Mode Collapse in Reinforcement Learning and How to Fix It with Inverse Probability Scaling (arxiv.org)

TTCS: Test-Time Curriculum Synthesis for Self-Evolving (arxiv.org)

Identifying Dependent Components from Multi-domain Linear Mixtures (arxiv.org)

Characterizing Cross-Relay Generalization Failure in Machine Learning-Based Power System Protection (arxiv.org)

Beyond the Point Mass: Acceptance-Derived Weights and Distributional Losses for Speculative Decoding (arxiv.org)

Relative Representations Enable Portable White-Box Monitoring Across Language Models (arxiv.org)

A general-purpose System One decision model for clinical records (arxiv.org)

Omni-Woodpecker: Hallucination Correction via Evidence-Prior Rebalancing for Omni LLMs (arxiv.org)

PROVISO: Provenance-Informed Action Selection for Secure LLM Agents (arxiv.org)

PORA: Learning When More Reasoning Pays Off for LLM Agents (arxiv.org)

EDMA: Entropy-Driven Multimodal Answering (arxiv.org)

Don't Skip the Baseline: A Controlled Comparison of Dynamic Expert Skipping in MoEs (arxiv.org)

Self-Credit: Credit Assignment for Long-Horizon Agentic RL via Intra-group Rescaling (arxiv.org)

Rewards for Budget-Aware Multi-Turn Agents That Adapt to Arbitrary Test-Time Constraints (arxiv.org)

Region-Specific EEG Classification of Hypnotic Susceptibility Using a Sliding-Window 2D CNN: A Subject-Independent Evaluation (arxiv.org)

MoSAIK: Motion Search with Artificial Intelligence in K-space (arxiv.org)

LSA-PAST: Layer-Specific Adaptation and Prototype Transport for Compositional Incremental Learning (arxiv.org)

Sparse RL Relaxation for LLM Reasoning (arxiv.org)

Magnolia: A Persistent-Memory Agent Framework for Computational Chemistry (arxiv.org)

CNN-based classifier for subject-independent MBTI Extraversion/Introversion preference classification from short-time EEG signals (arxiv.org)

Auditing Safety Preference Supervision with Recorded Derivations (arxiv.org)

What Would Removing This Corruption Change? Failure Attribution for Vision-Language-Action Models (arxiv.org)

A latent control model for realistic rodent motion (arxiv.org)

TEX: Test-Time Scaling Testing Agents via Execution-based Cross-Validation (arxiv.org)

EGI-BENCH: An Enterprise General Intelligence Benchmark Suite (arxiv.org)

A latent control model for realistic rodent motion (arxiv.org)

TEX: Test-Time Scaling Testing Agents via Execution-based Cross-Validation (arxiv.org)

Vulnerability Analysis for Safe Multi-Agent Reinforcement (arxiv.org)

Stronger artificial intelligence breakthroughs hasten their own exhaustion (arxiv.org)

What Not to Erase: Saliency-Constrained Augmentation for Low-Resource Handwritten Character Recognition (arxiv.org)

Physics-Inspired Optimizers beyond the Edge of Stability: Period-Two Orbits and Delayed Instability (arxiv.org)

EMBER: Predicting TCR--pMHC Binding Without a Docked Complex (arxiv.org)

Robust Lightweight IoT Device Fingerprinting Under Network Traffic Anomalies (arxiv.org)

DStill: Trajectory Consistency Distillation for Accelerated One-Step Drafting in Block Diffusion Speculative Decoding (arxiv.org)

When Does an Added Planning Capability Create Operational Value? An Observable Evidence Standard (arxiv.org)

Evolving Self-tuning Benchmarks for Long-Horizon Tool Use (arxiv.org)

Unsupervised Tool Evolution for Coding Agents in Data-Lake Analytics (arxiv.org)

Modeling Authentic Research Trajectories: A Graph View of Research Ideation (arxiv.org)

Understanding Data Augmentation under Label Noise (arxiv.org)

Training Controllable World Models via Action Influence Maximization (arxiv.org)

Steering intonational tunes in text-to-speech synthesis (arxiv.org)

Learning the Geometry of Materials Space with Trained Neural Tangent Kernels (arxiv.org)

The Router Already Knows (arxiv.org)

Baseline and Incremental: User, Item, and Request Debiasing in Instagram Feed Ranking (arxiv.org)

SCODEL: Shared Canonical Ontology Definitions for Food Entity Linking and Ontology-Grounded Recipe Knowledge Graphs (arxiv.org)

CUA-Recovery: Benchmarking Error Recovery in Long-Horizon Computer-Use Agents (arxiv.org)

Does Multi-Agent Collaboration Pay When Agents Serve Users? Testing Agent Scaling in Dual-Control Settings (arxiv.org)