AI/ML arXiv cs.AI

WebSwarm: Recursive Multi-Agent Orchestration for Deep-and-Wide Web Search

WebSwarm introduces a recursive multi-agent orchestration framework for deep and wide web search, improving coverage and depth over single-agent systems.

Other Hacker News

Einstein's relativity rules chemical bonds in heavy elements, new research shows

New research indicates that Einstein's theory of relativity plays a critical role in determining the chemical bonds of heavy elements.

AI/ML Hacker News

AI 2040: Plan A

A discussion or proposal regarding a long-term plan for AI development leading up to the year 2040.

Software Engineering Hacker News

After 7 years in production, Scarf has reluctantly moved away from Haskell

The company Scarf details its decision to move away from Haskell after seven years of production use, discussing the trade-offs involved.

AI/ML arXiv cs.AI

EgoWAM: World Action Models Beyond Pixels with In-the-Wild Egocentric Human Data

Introduction of EgoWAM, a framework for robot manipulation that uses World Action Models to better transfer human egocentric data to robots.

AI/ML arXiv cs.AI

ADORN: Adaptive Drift handling for Open RAN using Reinforcement Learning

Proposes ADORN, a Q-learning-based adaptive retraining approach to handle drift in Open Radio Access Networks (O-RAN) using RL.

AI/ML arXiv cs.AI

Spatio-Temporal Scheduling Prediction Under Backhaul Delay for Resilient Coordinated Beamforming

A two-stage predictive framework using StemGNN to handle backhaul delay in coordinated beamforming for 5G networks.

AI/ML arXiv cs.AI

Two Axes of LLM Abstention: Answer Correctness and Question Answerability

Research on LLM abstention, distinguishing between incorrect answers and unanswerable questions via hidden-state probes.

AI/ML arXiv cs.AI

VEGAS: Human-Aligned Video Caption Evaluation via Gaze

VEGAS is a training-free metric that uses human gaze data to evaluate and select personalized video captions.

AI/ML arXiv cs.AI

The Context Access Divide: Interaction-Level Architecture as a Complementary Dimension of Agentic Inequality

Analysis of the 'Context Access Divide', exploring how autonomous context retrieval versus manual attachment creates disparities in AI utility.

AI/ML arXiv cs.AI

Cognitive-structured Multimodal Agent for Multimodal Understanding, Generation, and Editing

A Cognitive-structured Multimodal Agent that uses episodic visual memory to improve long-horizon multimodal dialogue and reasoning.

AI/ML Hacker News

Inference Optimization for MiMo v2.5: Pushing Hybrid SWA Efficiency to the Limit

A technical discussion on optimizing inference for MiMo v2.5, focusing on enhancing the efficiency of Hybrid SWA.

Tech Business/VC Hacker News

Apple sues OpenAI, accuses ex-employees of stealing trade secrets

Apple is suing OpenAI, alleging that former employees stole trade secrets to facilitate the development of their models.

Cybersecurity Hacker News

GhostLock, a stack-UAF that has existed in ALL Linux distributions for 15 years

Discovery of 'GhostLock', a Use-After-Free (UAF) vulnerability in the Linux kernel stack that has persisted across all distributions for 15 years.

Software Engineering Hacker News

Postgres locks do not scale

A deep dive into why PostgreSQL locks do not scale, analyzing the internal locking mechanisms and their limitations.

Open Source Hacker News

Show HN: SubjectiveZero, an open-source agentic node editor for creative coding

Introduction of SubjectiveZero, an open-source agentic node editor designed for creative coding tasks.

Tech Business/VC TechCrunch

Bluesky’s interim CEO, Toni Schneider, drops the ‘interim’

Toni Schneider has transitioned from interim to permanent CEO of the social media platform Bluesky.

Other The Verge

The FCC is cracking down on DJI tech that dodged the foreign drone ban

The FCC is fining front companies used by DJI to bypass US drone bans.

Other Ars Technica

China recovered its first reusable rocket and showed a new way to do it

China has successfully recovered its first reusable rocket, marking a significant step in its space program's capabilities.

AI/ML arXiv cs.AI

When Synthetic Speech Is All You Have: Better Call GRPO

Research showing that Group Relative Policy Optimization (GRPO) significantly improves ASR adaptation using synthetic speech compared to SFT.