AI/ML arXiv cs.AI

When AI Classifies: What Counts as Public Administration?

A study analyzing how AI-assisted classification systems impact the visibility and intellectual structure of public administration scholarship.

AI/ML arXiv cs.AI

CHARLIE: An On-Premise Multi-Agent Retrieval-Augmented Generation System for Evidential Reasoning in Forensic Science

Introduction of Charlie, an on-premise multi-agent RAG system specifically designed for the high-security requirements of forensic science.

AI/ML arXiv cs.AI

Modality Relevance is not Modality Utility: Post-hoc Selective Modality Escalation for Cost-Aware Multimodal RAG

A research paper proposing post-hoc selective modality escalation to reduce the cost of multimodal RAG by only using expensive VLMs when necessary.

AI/ML arXiv cs.AI

PORTS: Preference-Optimized Retrievers for Tool Selection with Large Language Models

Introduction of PORTS, a preference-optimized retriever designed to improve tool selection accuracy for LLMs using a novel odds ratio optimization method.

AI/ML arXiv cs.AI

Scientific Code Search at Scale: A Multi-Domain Dataset and Benchmark

The release of a large-scale multi-domain dataset and benchmark for scientific code search, spanning various NASA Science Mission Directorate divisions.

AI/ML arXiv cs.AI

Geometry-Aware Infrastructure-Anchored Denoiser for UWB Sensing and Work-Zone Reconstruction

Introduction of GAIA, a geometry-aware learning framework for denoising UWB sensing data to improve work-zone reconstruction in intelligent transportation.

AI/ML arXiv cs.AI

The Granularity Paradox: How Temporal Disaggregation Inflates In-Sample Fit and Compounds Out-of-Sample Error

An investigation into the 'Granularity Paradox' in time-series forecasting, showing how high-frequency data can compound out-of-sample errors.

AI/ML arXiv cs.AI

Empirical Minimal-Realisation Compression of Deep Neural Networks via Controllability-Observability Tests

A new framework for compressing deep neural networks based on controllability-observability tests, achieving significant parameter reduction with minimal accuracy loss.

Tech Business/VC Hacker News

The space bit of SpaceX is worth $8 a share, says Morgan Stanley

Morgan Stanley estimates the space-related portion of SpaceX's valuation at approximately $8 per share.

AI/ML arXiv cs.AI

The Jagged Global Economy: Frontier AI Unevenly Exposes National Economies

Researchers introduce a national AI exposure metric to analyze how frontier AI unevenly affects labor markets across 141 countries, highlighting significant disparities between high- and low-income nations.

AI/ML arXiv cs.AI

CCBENCH: Assessing LLM Cultural Competence via Implicitly Signaled Norms using Health Queries

CCBENCH is a new framework for evaluating the cultural competence of LLMs, revealing that current models struggle to adapt to implicit cultural norms in health-related queries.

AI/ML arXiv cs.AI

A Guiding Framework for K-12 Teachers in Creating AI-powered Learning Technologies through Vibe Coding

The GAIDE framework provides K-12 teachers with structured guidance to create AI-powered learning tools using 'vibe coding' (natural language prompting).

Cybersecurity arXiv cs.AI

Position: Preventing AI-Generated CSAM Necessitates New Approaches to AI Safety

This paper argues that preventing AI-generated CSAM requires specialized AI safety approaches that can handle the legal and ethical constraints of child protection material.

Software Engineering arXiv cs.AI

Automated Recommendation of Programming Learning Content Using Pattern-based Knowledge Components

A new approach using pattern-based Knowledge Components (KCs) allows for the automated recommendation of introductory programming learning content based on semantic similarity.

AI/ML arXiv cs.AI

CANONIC: Governance Is Compilation

CANONIC proposes a governance framework that treats digital artifact admission into a corpus as a compilation process based on formal grammars and evidence ledgers.

AI/ML arXiv cs.AI

The GenAI Skill Bypass: Mapping Divergent Pathways of University Students and Staff AI Literacy

Research reveals a 'skill bypass' in university students where they master high-level GenAI creation tasks before acquiring foundational technical understanding of AI mechanics.

AI/ML arXiv cs.AI

Why does AI unlock new possibilities in STEM education? A Bibliometric Analysis of Trends and Future Agenda

A bibliometric analysis explores how AI has shifted STEM education from simple knowledge transmission to capability development through intelligent scaffolding.

Hardware/Chips arXiv cs.AI

Contrastive Predictive Coding with Compression for Enhanced Channel State Feedback in Wireless Networks

A unified compression-prediction framework using Contrastive Predictive Coding (CPC) is proposed to improve channel state feedback in 3GPP-compliant wireless networks.

Software Engineering Hacker News

Decoding the obfuscated bash script on a Uniqlo t-shirt

A technical breakdown of an obfuscated bash script found on a Uniqlo t-shirt.

Other Hacker News

Ants: Who looks after the injured in a colony?

A discussion about how injured ants are cared for within a colony.