AI News

Universal Defenses for Tool-Integrated LLM Agents Against Adversarial Attacks (arxiv.org)

AI-Driven Feedback Systems, Digital Labour, and Silent Quitting: Transforming African Workplaces (arxiv.org)

Permutation-Based Stegomalware in Large Language Models: Threats and Countermeasures (arxiv.org)

SceneBench: A Hierarchical Benchmark for Vision-Language Understanding of 3D Scenes (arxiv.org)

Models as Governed Interfaces for AI-Native MBSE: Read-Side Adequacy and Write-Side Admissibility (arxiv.org)

Protocol-Preserving Context Trimming for Agentic Workflows: Benefits, Failure Regimes, and Budget Guardrails (arxiv.org)

AI Policies: Help or Hindrance? A Software Developer's Perspective (arxiv.org)

Geospatial Metadata Improves Discoverability by Connecting Datasets Across Scientific Disciplines (arxiv.org)

Beyond the Name: Demographic Leakage in De-Identified R\'esum\'es and Evaluation Artifacts in LLM Bias Audits (arxiv.org)

Competence-Preserving Resume Perturbations Expose Presentation Sensitivity in LLM Screening (arxiv.org)

QALPA: Property-guided diffusion modeling for efficient exploration of chemical spaces of flexible molecules (arxiv.org)

What Does Layer-Importance Reveal About Transformers and State-Space Models? (arxiv.org)

"Looking for Something Weird to Happen": How Humans Sustain AI Agent Novelty Amid Semantic Collapse (arxiv.org)

Never Stop Thinking: Continuous-Time Language Agents (arxiv.org)

World Model Science: Self-Organized Criticality, Weak Chaos, and Metastable Belief Dynamics in Long-Horizon LLM Agents (arxiv.org)

Transformer-Based Token Fusion and Dynamic Graph Planning for Audio-Visual Navigation (arxiv.org)

Talking Head Synthesis with Facial Landmark Guidance via 3D Gaussian Splatting (arxiv.org)

JustFit: 200K-Token LLM Serving on a 24 GiB Laptop with Just-in-Time State Management (arxiv.org)

LimiX-2: A Contextual Mechanism Network Towards General Structured-Data Intelligence (arxiv.org)

RepoAtlas: Guiding Coding Agents via Evolving Multimodal Repository Views (arxiv.org)

Continuous-Time Machine Learning: A Unified Mathematical Perspective (arxiv.org)

A Systematic Evaluation of Machine Learning Methods for Fault Detection and Line Identification in Electrical Power Grids (arxiv.org)

TAME: Token Attribution and Masking for Emergent misalignment (arxiv.org)

Noise2Noise Revisited: Training Pair Distributions Dominate Loss Choice in Self-Supervised Denoising (arxiv.org)

SOTER: A Generative Time-Series Foundation Model for Wearable Human Physiological Signals (arxiv.org)

Geometry of learning dynamics: Gradient descent versus natural gradient on the ridge of optimization (arxiv.org)

ImpossibleRubrics: Stress-Testing Generated Rubrics as Reward Signals (arxiv.org)

FlowATC: Aircraft Trajectory Prediction via Flow Matching (arxiv.org)

On the Importance of Gating: Memorization vs. In-Context Learning in State Space Models (arxiv.org)

AsyncCouple-Flow: Asynchronous Cross-Modal Coupling and Flow Matching for Spatio-Temporal Forecasting (arxiv.org)

Recovering Physical Parameters from Fragmented Observations via Exact Distributed Spline Merging (arxiv.org)

Divergence Timing and Cumulative Disagreement under KV-Cache Eviction (arxiv.org)

GrowMTP: Can RL Grow Its Own Draft Head? (arxiv.org)

Right Direction, Wrong Step: Geometric Analysis of Finite-Step Failure in Looped Transformers (arxiv.org)

Single Document Extractive Summarization using Domination in Hypergraph (arxiv.org)

Repurposing Unified Topological Signatures for Graph Representation Learning (arxiv.org)

High-Fidelity Digital Twin Data Models by Randomized Dynamic Mode Decomposition and Deep Learning with Applications in Fluid Dynamics (arxiv.org)

Neural Field Ensembles for Aerodynamic Surface Prediction: Winning Solution to the ONERA CRM Wall Distribution 2025 Challenge (arxiv.org)

A unified framework for global and local interpretability using adaptive derivative-ordered random explanation (arxiv.org)

IRENE: A Convolutional GRU Ensemble Model for Radar Precipitation Nowcasting over Italy (arxiv.org)

LoopSpec: Pipelined Self-Speculative Decoding for Looped Transformers (arxiv.org)

MyoFlow: Anchor-Tied Rectified Flow for HD-sEMG Gesture Recognition Across Sessions and Subjects (arxiv.org)

Memorisation bias in medical AI (arxiv.org)

Latent Undertow: How Ordinary Typos Break Probes (arxiv.org)

Easy to Catch a Liar, Hard to Clear an Honest One: Language Models Diagnosing a Corrupted Reward Channel from a Verified Record (arxiv.org)

LCAP: Population-Informed Latent Chip Adaptation from Few Output Probes for Photonic Neural Networks (arxiv.org)

Same Flow, Different Paths: Variance Reduction in Flow Matching (arxiv.org)

Large Language Models Develop Belief State Geometry In-Context (arxiv.org)

OPEN-1B: A Fully Auditable Training Run (arxiv.org)

Bridging the Confidence Gap: Temperature Scaling for Calibrating Test-Time Prompt Tuning (arxiv.org)