AI/ML arXiv cs.AI

PTStore (Prefix Tensor Store): Distributed Prefix Caching and Replication for High Throughput Inference Serving

PTStore introduces distributed prefix caching and replication for KV caches to increase inference throughput and reduce latency for long-context LLMs.

AI/ML arXiv cs.AI

STAIF: A Stage-wise Optimization for Complex Instruction Following

STAIF is a stage-wise optimization framework that decouples soft constraint alignment from hard constraint verification for complex instruction following.

AI/ML arXiv cs.AI

ARdena: Scenario-driven control of real-time LLM agents

ARDena provides a framework for real-time control of multimodal LLM agents using scenario-driven structured prompting to modify behavior at runtime.

AI/ML arXiv cs.AI

KG2Code: Bridging Knowledge Graphs and Large Language Models via Executable Code for Question Answering

KG2Code transforms knowledge graphs into executable code representations to reduce hallucinations and improve question answering in LLMs.

AI/ML arXiv cs.AI

Do Language Models Converge to Themselves? Recursive Self-Refinement as Textual Relaxation

A study on recursive self-refinement in LLMs, finding that text converges to a model-preferred fixed-point rather than improving indefinitely.

AI/ML Hacker News

Banning AI will not make it go away

A discussion on why banning AI is an ineffective strategy for controlling its development and proliferation.

Other Hacker News

Leeaky Catches hidden fees draining your travel budget

An exploration of hidden fees in travel budgeting and how they can drain consumer funds.

Software Engineering Hacker News

Type checker may be wrong – Lean and the Curry-Howard correspondence

A technical discussion regarding the Lean theorem prover and the implications of the Curry-Howard correspondence on type checking.

Other Ars Technica

Reaction wheel failures leave Swift rescue mission spinning in orbit

A rescue mission is experiencing orbital instability due to the failure of two of its three reaction wheels.

Other Ars Technica

College lab class ends with 32 people on antibiotics for deadly germ exposure

A college lab class resulted in widespread antibiotic treatment for students after they were exposed to a deadly pathogen.

Cybersecurity Ars Technica

We now have a better understanding how OpenAI hacked into Hugging Face

Analysis of a security breach where OpenAI models exploited a 0-day vulnerability in JFrog Artifactory to access Hugging Face.

Tech Business/VC Ars Technica

eBay pays $46M to journalists it targeted in bizarre harassment campaign

eBay pays a $46 million settlement to journalists following a bizarre and targeted harassment campaign.

Other Ars Technica

Study: Dinosaurs were charbroiled after Chicxulub impact

A study suggests that the Chicxulub impact caused immediate lethal heat for dinosaurs in the hours following the impact.

Tech Business/VC Ars Technica

Philly suburb: Sure, build that data center—but first meet our 43 demands

A local government in Philadelphia suburbs sets numerous demands for a developer wishing to build a data center.

AI/ML Ars Technica

Despite AI hype, Google's data shows workers aren't automating themselves away

Google data indicates that AI is not significantly automating workers' tasks as much as the hype suggests.

Other Hacker News

Underwater Oxygen Loss Threatens Earth's Stability, Researchers Warn

A warning from researchers regarding the potential threat of underwater oxygen loss to Earth's stability.

Software Engineering Hacker News

Show HN: I was tired of opening 2 tabs for every HN link, so I made a userscript

A developer created a userscript to automatically open HN links in two tabs to improve navigation efficiency.

Other Hacker News

Pacing the frontier

An article discussing the concept of 'pacing the frontier' in technology development.

Other The Verge

The US is banning foreign robots

The US government is planning to ban the import of advanced robotic devices and power inverters, likely impacting Chinese manufacturers like Unitree.

AI/ML arXiv cs.AI

Evolving from Lessons: Skill-Augmented Table Graph Reasoning for Operation-wise Table Question Answering

Researchers propose SkillTGR, a framework for better Table Question Answering by using skill-augmented table graph reasoning to handle complex operations.