AI/ML arXiv cs.AI

From Sorting Algorithms to Scalable Kernels: Bayesian Optimization in High-Dimensional Permutation Spaces

A new framework for Bayesian Optimization in high-dimensional permutation spaces using the 'Merge Kernel' derived from merge sort to improve computational efficiency.

AI/ML arXiv cs.AI

Fusion of Pervasive RF Data with Spatial Images via Vision Transformers for Enhanced Mapping in Smart Cities

A deep learning approach combining DINOv2 and RF data to improve building mapping accuracy in smart cities.

AI/ML arXiv cs.AI

Tackling GNARLy Problems: Graph Neural Algorithmic Reasoning Reimagined through Reinforcement Learning

The GNARL framework reframes neural algorithmic reasoning as a reinforcement learning problem to solve combinatorial NP-hard graph problems without needing expert algorithms.

AI/ML arXiv cs.AI

Q-Net: Queue Length Estimation via Kalman-based Neural Networks

Q-Net is an AI-augmented Kalman filter framework for real-time queue length estimation at intersections using privacy-preserving vehicle counts and speed data.

AI/ML arXiv cs.AI

Shift-Invariant Attribute Scoring for Kolmogorov-Arnold Networks via Shapley Value

ShapKAN is a pruning framework for Kolmogorov-Arnold Networks (KANs) that uses Shapley value attribution to achieve shift-invariant node importance ranking for better compression.

Tech Business/VC Hacker News

Microsoft turns to AWS as GitHub faces AI capacity crunch

Microsoft is utilizing AWS infrastructure to supplement its AI capacity needs as GitHub faces a shortage of compute resources.

Tech Business/VC The Verge

Inside the fight over Claude Mythos 5

Anthropic is facing a legal battle with the US government over export control directives that restrict access to its Mythos 5 and Fable 5 AI models.

AI/ML arXiv cs.AI

MirrorCheck: Efficient Adversarial Defense for Vision-Language Models

MirrorCheck is a new model-agnostic detection framework designed to protect Vision-Language Models from adversarial attacks using semantic consistency checks.

AI/ML arXiv cs.AI

Vanishing Depth: Training Generalized Depth Adapters with Sinusoidal Depth Preprocessing for Pretrained RGB Encoders

Researchers propose a self-supervised depth adapter with sinusoidal preprocessing to improve metric depth understanding in pretrained RGB encoders for robotics.

AI/ML arXiv cs.AI

Revisiting Outage for Edge Inference Systems

A new theoretical framework introduces 'inference outage' (InfOut) to optimize the tradeoff between communication overhead and reliability for 6G edge inference systems.

Hardware/Chips arXiv cs.AI

FPGA-Based Neural Network Accelerators for Space Applications: A Survey

A comprehensive survey on using FPGA-based neural network accelerators to enhance onboard computing systems for space missions.

AI/ML arXiv cs.AI

UniversalRAG: Retrieval-Augmented Generation over Corpora of Diverse Modalities and Granularities

UniversalRAG is an any-to-any Retrieval-Augmented Generation framework that integrates knowledge from heterogeneous sources across diverse modalities and granularities.

Cybersecurity arXiv cs.AI

The Accountability Paradox: How Platform API Restrictions Undermine AI Transparency Mandates

An analysis of how platform API restrictions on social media hinder algorithmic transparency and compliance with the EU Digital Services Act.

AI/ML arXiv cs.AI

Fractured Chain-of-Thought Reasoning

Fractured Sampling is introduced as a way to reduce token costs in Chain-of-Thought reasoning by truncating reasoning traces without significant loss in accuracy.

Open Source arXiv cs.AI

MASLab: A Unified and Comprehensive Codebase for LLM-based Multi-Agent Systems

MASLab is a unified codebase that integrates over 20 LLM-based multi-agent system methods to provide a standardized environment for research and evaluation.

Tech Business/VC Hacker News

Amazon Announces Multibillion-Dollar Data Center in Missouri

Amazon is building a multibillion-dollar data center in Missouri.

AI/ML Hacker News

Cohere's First Model for Developers

Cohere releases its first AI model specifically tailored for developers.

Software Engineering Hacker News

Reviews have become expensive, rewrites have become cheap

An exploration of the shifting cost dynamics between reviewing existing code versus rewriting it, especially in the context of AI.

AI/ML Hacker News

Humanity isn't ready for the coming intelligence explosion

A speculative discussion on humanity's readiness for a potential intelligence explosion.

AI/ML Hacker News

The 90-year-old idea behind JEPA models: Canonical Correlation Analysis

An analysis of the role of Canonical Correlation Analysis in the development of JEPA models.