AI/ML arXiv cs.AI

Large Language Models for Token-Efficient and Semantic-Preserving Opinion Summarization

A framework for token-efficient and semantic-preserving opinion summarization using stratified sampling.

AI/ML arXiv cs.AI

3D-DefectBench: A Controlled Factorial Study of Vision-Language Model Evaluation Pipelines for Fine-Grained 3D Generation Defects

Introduces 3D-DefectBench, a benchmark for analyzing vision-language model pipelines for detecting 3D generation defects.

Other Hacker News

QR-Swastika-Avoider

A tool designed to prevent the generation or scanning of QR codes containing offensive imagery, specifically swastikas.

Other Hacker News

Microsoft deleted my account and OneDrive

A user reports the sudden deletion of their Microsoft account and associated OneDrive data without clear explanation.

Other Hacker News

Mathematical texts from a Maya site in Guatemala identify an ancient astronomer

Ancient mathematical texts from a Maya site in Guatemala provide new insights into the identity of an ancient astronomer.

AI/ML Hacker News

LeMario: Training a JEPA World Model on Super Mario Bros

LeMario project focuses on training a Joint Embedding Predictive Architecture (JEPA) World Model on Super Mario Bros games.

Other TechCrunch

Lorde says AI glasses are “not sexy”

Singer Lorde expresses skepticism and lack of attraction towards AI-powered smart glasses.

Cybersecurity Ars Technica

Microsoft’s Secure Boot has been broken for a decade and no one noticed until now

A critical decade-long vulnerability in Microsoft's Secure Boot has been discovered, allowing simple bypasses via unrevoked shims.

Other Ars Technica

Trump admin puts Americans in Congo on "do-not-board" list, barring return

The Trump administration has implemented a 'do-not-board' list for Americans in Congo, requiring a 21-day quarantine in a third country.

AI/ML arXiv cs.AI

Multi-Scale Convolution with Optimal Transport Attention Effect on Multivariate Time Series

Introduction of MSC-OT, a new architecture for multivariate time series forecasting using multi-scale convolution and optimal transport attention.

Software Engineering arXiv cs.AI

Lightning Fast Matching Dependency Discovery with Desbordante

The Desbordante profiler implements optimized HyMD algorithms for significantly faster discovery of matching dependencies in data quality tasks.

AI/ML arXiv cs.AI

LSTrans: Efficient Knowledge Transfer for Lightweight and Automated ECG Classification

LSTrans is a lightweight hybrid model combining 1D CNNs and Transformers for efficient ECG classification on wearable devices.

Tech Business/VC Hacker News

Financing the AI boom: from cash flows to debt [pdf]

An analysis of the shift in AI boom financing from traditional cash flows to debt-based models.

Other Hacker News

Microsoft Deletes User's 25-Year-Old Account with Thousands Spent on Games

A report on Microsoft deleting a long-term user's account, resulting in the loss of significant financial investment in games.

Other Hacker News

Banter

A Hacker News thread discussing 'Banter'.

Hardware/Chips The Verge

Nothing’s good-looking Watch 3 Pro smartwatch is just $69

The CMF by Nothing Watch 3 Pro is a budget-friendly smartwatch featuring an OLED display and long battery life, priced at $69.

AI/ML arXiv cs.AI

Action Map Policy: Learning 3D Closed-loop Manipulation via Pixel Classification

Introduction of Action Map Policy (AMP), a method that treats 3D closed-loop robot manipulation as a pixel classification problem to improve precision and inference speed.

AI/ML arXiv cs.AI

MDQEC-QAS: Meta-Decoding for Quantum Error Correction with Hardware-Aware VQC Search and Confidence-Gated Recovery

A meta-decoding framework for quantum error correction using hardware-aware variational quantum circuits (VQC) and confidence-gated recovery.

AI/ML arXiv cs.AI

PromptGraph: Graph-Guided Prompt Sanitization for Balancing Privacy and Utility in LLM Inference

PromptGraph is proposed as a graph-guided prompt-sanitization approach to balance privacy and utility during LLM inference.

Cybersecurity arXiv cs.AI

Distributed Denial of Science: How Indirect Data Poisoning of AI Systems Can Industrialize Scientific Fraud

Research on 'Distributed Denial of Science', exploring how indirect data poisoning of AI systems can be used to industrialize scientific fraud.