AI/ML TechCrunch

Open-weight AI models are catching up to the frontier. The safety gap remains. 

A report indicates that open-weight AI models like GLM-5.2 are nearing frontier capabilities but still lack critical safety mitigations.

Tech Business/VC TechCrunch

Anthropic signs $10B deal with AI cloud startup Volta

Anthropic has entered into a reported $10 billion cloud partnership deal with AI cloud startup Volta.

AI/ML arXiv cs.AI

StaQ: a Finite Memory Approach to Discrete Action Policy Mirror Descent

Researchers propose StaQ, a finite memory approach to Discrete Action Policy Mirror Descent in RL to reduce memory requirements while maintaining performance.

AI/ML arXiv cs.AI

Towards White-Box Deep Wireless Sensing

RF-CRATE is introduced as a white-box deep wireless sensing model that uses complex-valued transformers for better interpretability and reliability.

AI/ML arXiv cs.AI

Patch-Based 3D Variational Autoencoder for Super-Resolution of Turbulent Channel Flow

A patch-based 3D Variational Autoencoder is proposed for the super-resolution of turbulent channel flow, outperforming traditional interpolation methods.

AI/ML arXiv cs.AI

RePaCA: Leveraging Reasoning Large Language Models for Static Automated Patch Correctness Assessment

RePaCA is a new static automated patch correctness assessment technique using reasoning LLMs and RL to identify overfitting patches in software repair.

AI/ML arXiv cs.AI

"Not in My Backyard": LLMs Uncover Online and Offline Social Biases Against Homelessness

Researchers use LLMs to analyze social biases against homelessness, releasing a multi-domain PEH bias corpus and an audit protocol.

AI/ML arXiv cs.AI

Adaptive Policy Backbone via Shared Network

The Adaptive Policy Backbone (APB) method is proposed for meta-transfer RL to improve sample efficiency and adaptation to out-of-distribution tasks.

AI/ML arXiv cs.AI

Fast Feature Field ($\text{F}^3$): A Predictive Representation of Events

Fast Feature Field (F3) is a new predictive representation for event-based cameras that achieves state-of-the-art performance in robotic vision tasks.

AI/ML arXiv cs.AI

Epistemic-aware Vision-Language Foundation Model for Fetal Ultrasound Interpretation

FetalMind is a medical AI system for fetal ultrasound interpretation that utilizes a large-scale report corpus and an epistemic-aware design.

Other Hacker News

Waymo – Dallas Open to All

Waymo's ride-hailing service is expanding its availability to all users in Dallas.

Tech Business/VC Hacker News

Investors in Situational Awareness deserved to lose their shirts

A discussion on Hacker News regarding investors in the 'Situational Awareness' AI safety concept losing money.

Other Hacker News

Stephen Wolfram's Wife Has Died

Announcement regarding the death of Stephen Wolfram's wife.

Other Hacker News

The Red Strings Club

Discussion on The Red Strings Club, likely a game or conceptual project.

AI/ML TechCrunch

Meet Wrinkles, an AI app that uncovers the hidden stories of the places around you

Wrinkles is a new AI-powered audio tour guide app that reveals local histories of places.

Cybersecurity TechCrunch

Nvidia doesn’t mess around: A week after open AI industry group formed, it’s already showing progress

Nvidia is leading the Open Secure AI Alliance to create standards for defending against AI agents.

AI/ML arXiv cs.AI

Revisiting Multi-Permutation Equivariance through the Lens of Irreducible Representations

This research paper proposes a methodology using irreducible representations and Schur's lemma to characterize equivariant linear layers for permutations.

AI/ML arXiv cs.AI

Dual-Force: Enhanced Offline Diversity Maximization under Imitation Constraints

Dual-Force is an offline algorithm that maximizes diversity in behavioral policies using a Van der Waals force objective and Functional Reward Encoding.

AI/ML arXiv cs.AI

Dimensionality reduction for homological stability and global structure preservation

DiRe is a force-directed dimensionality reduction framework designed to preserve global structure and homological features.

AI/ML arXiv cs.AI

Reproducing Human Individual Motor Signatures: A Data-Driven Approach for Repetitive Motion

A data-driven approach using LSTM neural networks to reproduce individual human motor signatures for realistic motion in avatars and robots.