hardware

🧵 Slime: The Most Elegant & Comfortable RL Training Framework Ever A deep dive into why Slime redefines LLM RL training with clean architecture & production-grade engineering ✨ Insights from Zhihu con

arXiv:2602.06283v2 Announce Type: replace Abstract: Exploiting sparsity during long-context inference is key to scaling large language models, as attention dominates the cost of autoregressive decodin

arXiv:2605.06763v1 Announce Type: new Abstract: Sparse attention improves LLM inference efficiency by selecting a subset of key-value entries, but at the cost of potential accuracy degradation. In par

arXiv:2603.23198v2 Announce Type: replace-cross Abstract: Scaling autoregressive large language models (LLMs) has driven unprecedented progress but comes with vast computational costs. In this work, w

arXiv:2509.05276v4 Announce Type: replace-cross Abstract: Mainstream Transformer-based large language models face major efficiency bottlenecks: training computation scales quadratically with sequence

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. A few months before he was awarded the Nobel Prize in economic

arXiv:2605.07655v1 Announce Type: cross Abstract: Searching a multi-biometric database of a billion records for a country-level identity system requires pushing the limits of all aspects of a biometri

arXiv:2508.02001v2 Announce Type: replace-cross Abstract: Pervasive encryption makes large-scale labeling infeasible for traffic analysis, while security operations demand edge analysis to avert servi

arXiv:2605.06927v1 Announce Type: cross Abstract: Object detection on heterogeneous edge devices must satisfy strict energy, latency, and memory constraints while still providing reliable perception f

arXiv:2605.07378v1 Announce Type: new Abstract: Zero-shot proxies, also known as training-free metrics, are widely adopted to reduce the computational overhead in neural network evaluation for scenari

This article likely examines the relationship between Arm Holdings (the British semiconductor design company), the UK government, and Apple, possibly covering topics such as Apple's use of Arm-based c

I was talking to a room of senior accountants a couple months ago and 10% had OpenClaw installations. Of course there are far more non-users and firms lag behind their people, but there is a sort of S

The market is trying to price a transition it hasn’t fully internalized. It sees Nvidia Corp.’s market cap with a five-handle and assumes the valuation is too high to grow further. We believe that’s t

“You are entering the world at an extraordinary moment,” NVIDIA founder and CEO Jensen Huang told graduates as he delivered the keynote address at Carnegie Mellon University’s 128th commencement cerem

Yann LeCun closed 1.03B for AMI Labs on March 10. Three days later, this paper dropped from his NYU collaborators. 15M parameters. Single GPU. A few hours of training. LeWorldModel is the first JEPA t

Neocloud company IREN Ltd. has secured a 2.1 billion commitment from the chipmaker Nvidia Corp. as part of a new data center partnership aimed at artificial intelligence workloads. The partners plan t

Learn how to deploy any Hugging Face model in one session using Goose and Together's Dedicated Container Inference. Skip the setup complexity — one prompt gets your model running in a production-grade

Great collab with @SakanaAILabs on an #ICML26 paper about sparse transformer kernels + formats optimized for modern NVIDIA GPU execution. • TwELL sparse packing • Fused CUDA kernels • 20%+ inference/t

Grammar-constrained decoding modifies language model generation by applying grammar constraints at each step to block structurally invalid tokens , ensuring syntactically correct Bash command generati

OPINION: Ever since private equity bought the @Veritasium YouTube channel, the content just hasn't been as good. Used to be a GREAT channel, maybe the best Science channel on YouTube, and now? SAD! Ve

Tickets are now available for the 4th annual AI Film Festival: June 11 at Alice Tully Hall at Lincoln Center in NYC, and June 18 at The Broad Stage in LA. It’s the biggest and most important celebrati

We’ve rolled out a significant update to Google Kubernetes Engine (GKE) that solves one of the most annoying problems in cloud infrastructure: cold start latency. GKE now has up to 4x faster node star

arXiv:2605.04074v1 Announce Type: new Abstract: AI data centers experience rapid fluctuations in power demand due to the heterogeneity of computational tasks that they have to support. For example, th

arXiv:2605.04595v1 Announce Type: new Abstract: The rapid adoption of large language models (LLMs) has created significant challenges for efficient inference at scale. Unlike traditional workloads, LL

This article discusses optimization techniques for maximizing efficiency on NVIDIA's GB200 NVL72 system using Slurm block scheduling, a job scheduling approach designed to improve resource utilization

arXiv:2605.04711v1 Announce Type: cross Abstract: Optimizer states occupy massive GPU memory in large-scale model training. However, gradients in different network blocks exhibit distinct behaviors, s

arXiv:2605.04357v1 Announce Type: cross Abstract: The usage of large language models (LLMs) has grown increasingly fragmented, with no single model dominating. Meanwhile, cloud providers offer a wide

arXiv:2605.05023v1 Announce Type: new Abstract: Efficient CUDA implementations of attention mechanisms are critical to modern deep learning systems, yet supporting diverse and evolving attention varia

arXiv:2605.04198v1 Announce Type: new Abstract: Performance of deep learning models is strongly governed by architectural capacity, with width and depth as primary controls. However, in physical-scien

arXiv:2605.04997v1 Announce Type: new Abstract: DualTCN is the first deep-learning framework for inverting time-domain marine controlled-source electromagnetic (MCSEM) transient data. Moving away from

Hot take on Elon’s surprise decision to rent 30 megawatts of compute to Anthropic: 1. It’s a tacit concession that xAI is not all that close to AGI (despite what he suggested last year). 2. It’s more

.@huggingface's agentic robotics app store for Reachy Mini is a big step toward more accessible physical AI. 🙌 Excited to see NVIDIA Isaac GR00T N integrated with Hugging Face LeRobot, helping develop

This post discusses a request for filtering functionality on Google Maps that would allow users to search for restaurants by ethnic cuisine type or authenticity level, suggesting a desire for better t

Lambda Labs secured a $1 billion senior secured credit facility to fund the expansion of its AI infrastructure and meet growing demand for gigawatt-scale computational resources. This financing enable

Less typing, more tanking. Faster logins mean more time in the gaming action — and this week provides GeForce NOW members with a smoother path straight into the battlefield. Cloud gaming is all about

arXiv:2604.01342v2 Announce Type: replace Abstract: Multivariate Hawkes processes are a widely used class of self-exciting point processes, but maximum likelihood estimation naively scales as O(N^2) i

Zheping Huang / Bloomberg: Moonshot, the Chinese AI startup behind Kimi chatbot, raised ~2B at a 20B+ valuation led by Meituan's venture arm; Moonshot's ARR topped 200M in April 2026 — Moonshot AI has

Jonathan Vanian / CNBC: Nvidia and data center operator IREN announce a deal to deploy up to 5 GW of AI infrastructure; Nvidia can invest $2.1B into IREN; IREN jumps 9%+ after hours — IREN shares surg

Nvidia's stock rose 2.6% following news that xAI is selling 220,000 used GPUs on the secondhand market, potentially indicating a shift in AI hardware demand or xAI's operational priorities. The report

arXiv:2605.03241v1 Announce Type: cross Abstract: Read-only memory (ROM) provides deterministic access to predefined data mappings. Extending ROM concepts to the optical domain enables high-bandwidth,

arXiv:2605.05049v1 Announce Type: cross Abstract: Frontier models increasingly adopt Mixture-of-Experts (MoE) architectures to achieve large-model performance at reduced cost. However, training MoE mo

AI will help build the energy it needs. That’s the case U.S. Energy Secretary Chris Wright and NVIDIA Vice President of Hyperscale and High-Performance Computing Ian Buck made Thursday morning at the

arXiv:2605.04497v1 Announce Type: new Abstract: Shapley values are a standard tool for explaining predictions of tree ensembles, with Path-Dependent SHAP being the most widely used variant. Despite su

Quantum computer maker Quantum Motion Ltd. today announced that it has raised 160 million in funding to enhance its silicon-based qubit technology. The Series C round was led by DCVC and Kembara. It c

NCCL Inspector is a profiling plugin that provides detailed, per-communicator, per-collective performance and metadata logging, designed to help users analyze and debug NCCL collective operations by g

In this post, you will learn how to secure reserved GPU capacity for short-term workloads using Amazon Elastic Compute Cloud (Amazon EC2) Capacity Blocks for ML and Amazon SageMaker training plans. Th