AI/ML arXiv cs.AI

SenseWalk: Agent-Based Semantic Trajectory Simulation Powered by Large Language Models in Zoned Environments

SenseWalk is an LLM-powered interactive system for simulating semantic trajectories in zoned environments to better understand human movement patterns.

Software Engineering arXiv cs.AI

SWE-Doctor: Guiding Software Engineering Agents with Runtime Diagnosis from Multi-Faceted Bug Reproduction Tests

SWE-Doctor is a software engineering agent that uses multi-faceted bug reproduction tests and runtime diagnoses to significantly improve automated patch generation.

AI/ML Hacker News

Memora: A Harmonic Memory Representation Balancing Abstraction and Specificity

A discussion regarding 'Memora', a proposed harmonic memory representation designed to balance abstraction and specificity in information storage.

Other Hacker News

U.S. Officials Believed Israel Was Plotting to Kill Iranian Negotiators

Reports on U.S. officials' beliefs regarding Israeli plots to target Iranian negotiators.

Tech Business/VC TechCrunch

Private space pilots are flying orbital missions for the US Space Force

Private space companies True Anomaly and Rocket Lab are conducting orbital satellite missions for the U.S. Space Force.

Other The Verge

Tesla driver faces manslaughter charges over Texas crash that killed a woman inside her home

A Tesla driver faces manslaughter charges after a fatal crash while allegedly using the Full-Self Driving (FSD) system.

AI/ML arXiv cs.AI

From World Models to World Action Models: A Concise Tutorial for Robotics

A tutorial explaining the design space of World Models and the introduction of World Action Models for robotics.

Cybersecurity arXiv cs.AI

Recovering Input Text from Hidden States: Study of Gradient-Based Inversion of Decoder-Only Language Models

Research on recovering original input text from the hidden states of decoder-only language models using gradient-based inversion.

AI/ML arXiv cs.AI

Meta-Transfer Learning for mmWave Beam Alignment

Introduction of MTL-BA, a meta-transfer learning framework that uses lightweight adapters to optimize beam alignment in mmWave MISO systems.

AI/ML arXiv cs.AI

CAT: Confidence-Adaptive Thinking for Efficient Reasoning of Large Reasoning Models

Proposed 'Confidence-Adaptive Thinking' (CAT) framework to reduce overthinking in Large Reasoning Models by modulating reasoning length based on self-certainty.

AI/ML arXiv cs.AI

Improving Sparse-View 3DGS Generalization via Flat Minima Optimization

A method to improve 3D Gaussian Splatting generalization in sparse-view scenarios using flat minima optimization and periodic reinitialization.

AI/ML arXiv cs.AI

DeWorldSG: Depth-Aware 3D Semantic Scene Graph Generation via World-Model Priors

DeWorldSG, a framework for generating 3D Semantic Scene Graphs from RGB-D sequences using world-model priors for robotics and AR.

Software Engineering Hacker News

Simple, beautiful Emacs modeline: modusregel

The article introduces modusregel, a simple and aesthetically pleasing modeline for the Emacs editor.

Cybersecurity Ars Technica

Newly discovered PamStealer isn't your typical macOS malware

Researchers have discovered PamStealer, a new type of macOS malware specifically designed to steal information from Mac users.

AI/ML arXiv cs.AI

Phantom References: Hallucinated Citations That Survive Peer Review at Top-Tier Conferences

A study reveals that hallucinated citations are appearing in peer-reviewed papers at top AI conferences, and introduces RefChecker to audit them.

AI/ML arXiv cs.AI

Prototype Memory-Guided Training-Free Anomaly Classification and Localization in Prenatal Ultrasound

A training-free framework for prenatal ultrasound anomaly classification and localization is proposed using prototype memory-guided learning.

AI/ML arXiv cs.AI

GaussianFusion: Unified 3D Gaussian Representation for Multi-Modal Fusion Perception

GaussianFusion introduces a unified 3D Gaussian representation for multi-modal fusion perception, improving 3D object detection and semantic occupancy.

AI/ML arXiv cs.AI

Active Learning for Cascaded Object Detection: Balancing Coverage and Uncertainty in Table Extraction Pipelines

The paper explores active learning for cascaded object detection in table extraction, proposing RankFusion and CAPA to balance coverage and uncertainty.

AI/ML arXiv cs.AI

LeVLJEPA: End-to-End Vision-Language Pretraining Without Negatives

LeVLJEPA is introduced as the first fully non-contrastive end-to-end vision-language pretraining method, providing stronger dense semantic features.

Software Engineering arXiv cs.AI

LRAT-Catcher: Importing SAT Solver Certificates into Lean4 by Reflection

LRAT-Catcher is a tool that imports SAT solver certificates into Lean 4, allowing for the formal verification of complex combinatorial problems.