AI News

Post-Training Quantization of Autoregressive Weather Models (arxiv.org)

Student-Guided Teacher Distillation for Efficient LLM Task Routing: Positioning Against Jev-Style System-1 Classifiers (arxiv.org)

Learning the Latent Structure: A Feature-Centric Approach to Graph Data Augmentation (arxiv.org)

Instance-Dependent Regret for CMDPs with Step-Wise Constraints (arxiv.org)

BaCP: Backbone Contrastive Pruning for Preserving Representations in Extremely Sparse Neural Networks (arxiv.org)

Beyond Predefined Sinks: Security-Aware Dependency Analysis for LLM Agents (arxiv.org)

OmniAct3D: Leveraging Foundation Geometry and Evidence-Grounded Reasoning for Panoramic 3D Detection (arxiv.org)

Tailoring the Quantization Space for 1-Bit KV Cache Compression (arxiv.org)

WebFovea: When the Model Is Right but the Click Is Wrong -- Reliable Round Trips for Vision-Based Web Agents on Live Websites (arxiv.org)

NegT2IBench: When Negation Changes the Picture. A Polarity Benchmark for Text-to-Image Models (arxiv.org)

Test-time Multi-agent Coordination by Decomposed Value Gradient Flow (arxiv.org)

Neuron merging via inverse-activation regression for post-training compression of sigmoid neural networks (arxiv.org)

OpenGameEval: Benchmarking Agentic Programming and Exploration in a Stateful Game Engine (arxiv.org)

DISSOLVR: An Interpretable and Fast Framework for Aqueous and Organic Solubility Prediction (arxiv.org)

What Is Lost in Post-Training? Default Collapse and the Loss of In-Context Steerability Across Diverse Perspectives (arxiv.org)

Zephon: Elastic Determinism for Online, Stateful Foundation Model Data Loading Pipelines (arxiv.org)

Securing Computer-Use Agents Against Branch Steering Attacks (arxiv.org)

ULTRADISCOVERY: Abductive Exploration in an Interconnected, Epistemically Open Universe (arxiv.org)

Beyond Single Videos: Benchmarking and Active Evidence Seeking for E-Commerce Cross-Video Reasoning (arxiv.org)

How Causality Bridges the Semantic Gap (arxiv.org)

Activation Sparsity with Weight Approximation for Faster LLM Decoding on Offloaded Weights (arxiv.org)

Seer: Maximum Likelihood Regression for Learning-Speed Curves (arxiv.org)

Scale-Recursive Rectified Flows for Few-Step Precipitation Ensembles (arxiv.org)

Ask, Relax, or Act? Evaluating Actionable Indeterminacy in LLM Preference Reasoning (arxiv.org)

How to Find and Reuse Policies for Continuous Adaptation in Lifelong Reinforcement Learning (arxiv.org)

Foresight: planning future perception in streaming VLMs without retraining (arxiv.org)

The Fragility of Trigger-Tag Mechanisms for Misuse Detection in Open-Weight LLMs (arxiv.org)

Online Verification of Language Model Responses Under Cost Constraints (arxiv.org)

Context-Tower Conversion Preserves Generation While Freezing Retains Knowledge: Low-Budget AR-to-Diffusion Conversion of MoE LLMs (arxiv.org)

AIGS: Adaptive Incremental Gating System for Online Representation Learning in Non-Stationary Data Streams (arxiv.org)

Mind the Refinement Gap: When Safe High-Level Robot Plans Produce Unsafe Executions (arxiv.org)

RAOA: Alternating-Operator Neural Computation with Programmable Radio Propagation (arxiv.org)

Constraint-Aware Training (arxiv.org)

EvoRiskBench: An Evolving Benchmark for Runtime Security Risks in Workspace Agents (arxiv.org)

Multimodal reasoning for broadly neutralizing antibody discovery from label-free human B cell repertoires across virus families (arxiv.org)

Evolving Hybrid Quantum-Classical Architectures for Image Classification (arxiv.org)

Predicting Steering Vectors and Adapter Weights for Few-Shot Author-Style Transfer (arxiv.org)

LiBRA: Detection-Aware Image Watermark Removal via Bidirectional Latent Optimization (arxiv.org)

Not Until the Evidence Says So: Teaching LLM Investigators When to Close a Case (arxiv.org)

Evaluation of Sampling Strategies and Physics-Informed Kolmogorov--Arnold Networks in Unbounded Domains (arxiv.org)

Demystifying LLM-as-a-Judge: Analytically Tractable Model for Inference-Time Scaling (arxiv.org)

ANCRe: Adaptive Neural Connection Reassignment for Efficient Depth Scaling (arxiv.org)

Performance Evaluation of Ising and QUBO Variable Encodings in Boltzmann Machine Learning (arxiv.org)

MobiAgent: Dual-Loop Recursive Policy Self-Improvement for Long-Horizon Mobile Manipulation (arxiv.org)

Detect and Suppress: A Mechanistic Defense against Adversarial Patches in VLA Models (arxiv.org)

Revisiting Input Time-frequency Representations in Multi-pitch Estimation for Vocal Ensembles (arxiv.org)

Weave Forcing: Compositional Memory Routing for Interactive Long Video Generation (arxiv.org)

Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation (arxiv.org)

Rethinking What to Cache in Few-Step Diffusion Transformers: Solver-Aware Target Selection (arxiv.org)

Threat-Preserving Representation Sensitivity in Agent-Security Benchmarks (arxiv.org)