Knowledge catalogue
This post speculates about a potential GTA 6 DLC called 'Scili Valley' based on leaked information, with the author expressing enthusiasm about its potential as a gaming experience. The post appears t
MaineCoon is the first video model that focuses on social interactions: facial expressions, emotions, fluid conversation, audio-lip sync, etc. Really impressive inference specs: 22B params, 47.5 FPS o
This handbook covers the technical aspects of running large language models efficiently at scale, focusing on GPU optimization techniques including GPU internals, key-value (KV) cache management, batc
Hot take: GLM 5.2 might be the first open/public model that actually changes the enterprise AI cost equation. I played with it for a few hours today after friends told me to stop ignoring it. I expect
This post likely presents a humorous or satirical perspective on experiencing computing systems based on the von Neumann architecture during a hypothetical memory shortage in 2026, possibly commenting
Speculation Is All You Need. In this blog post, we announce the co-release (w/ Z Lab) of six more state-of-the-art DFlash speculators for @Alibaba_Qwen 3.x. Over 1k output tps for 3.5 122B-A10B on a B
Clients of SemiAnalysis Memory and Accelerator model knew about Rubin Ultra 16 Hi to 12 Hi cut in March :) LMAO Rubin Ultra's HBM literally got downgraded to 12 Hi, and they are not even using HB yet
arXiv:2606.11390v1 Announce Type: new Abstract: Gaussian splatting methods have become increasingly popular for neural reconstruction of the real world. However, they are often limited in scale and re
arXiv:2606.11793v1 Announce Type: cross Abstract: Uncertainty in the terrestrial carbon cycle remains a major constraint in climate projections, partly driven by the uncertainties affecting the land s
arXiv:2606.11916v1 Announce Type: cross Abstract: This paper proposes an empirical methodology to study software aging in GPU-based LLM serving systems. Traditional aging studies focus on CPU-centric
Cloud-based artificial intelligence infrastructure startup TensorWave Inc. said today it has closed on a bumper 350 million Series B funding round as it strives to meet demand for an alternative to Nv
arXiv:2606.12340v1 Announce Type: new Abstract: Memory is not merely the storage of data; it is the scaffolding of reality. When biological memory fades, the world does not simply turn black; it regre
arXiv:2606.11381v1 Announce Type: new Abstract: Robotic strawberry harvesting requires precise 6D pose estimation; however, collecting 6D pose ground truth in real agricultural fields is inherently ch
arXiv:2606.11440v1 Announce Type: new Abstract: Existing multi-agent LLM orchestration methods, ranging from brute-force ensembles to learned routers, select models and topologies based on task and mo
arXiv:2605.06485v2 Announce Type: replace-cross Abstract: Large language models (LLMs) have transformed artificial intelligence, but their computational requirements remain prohibitive for most users.
arXiv:2512.22219v2 Announce Type: replace-cross Abstract: We introduce Mirage Persistent Kernel (MPK), the first compiler and runtime system that automatically transforms multi-GPU model inference int
arXiv:2606.11387v1 Announce Type: cross Abstract: Short pretraining runs can reduce experimental cost, but they can also over-promote configurations that only look strong at tiny budgets. We study an
This tweet from 14 hours ago is on track to get about a million views. But here’s the thing: the conclusion is true, but the tweet itself is already outdated. WSJ’s scoop that OpenAI is considering dr
arXiv:2606.10660v1 Announce Type: cross Abstract: AI inference services -- API subscriptions, enterprise chat tools, and SaaS products with embedded AI features -- fall unambiguously within Scope 3 Ca
Sarah Guo: As AI commoditizes benchmarkable work, an organization's lasting moats lie in tasks that are verifiable through its private data and judgment — The mid-2026 investor's version of AI psychos
arXiv:2606.10440v1 Announce Type: cross Abstract: Distributed machine learning (ML) is a key paradigm for today's large-scale artificial intelligence applications. As model inference arises as an impo
Amazon Web Services Inc.’s next-generation custom silicon is finally being made accessible to customers for the first time with the launch of the Elastic Compute Cloud M9g and M9gd instances. They’re
Battery energy storage systems serve as grid-interactive control assets that buffer fast-changing, power-dense AI loads, improve power quality, and enable flexible interconnection with utilities and d
arXiv:2606.10896v1 Announce Type: new Abstract: We present extbf{Flash-GMM}, a fused Triton kernel for efficient computation of Gaussian Mixture Models (GMMs) over large-scale data in a single GPU pas
This article recaps presentations from Ray Day NYC featuring case studies from Discord, Cubist, and Coinbase on scaling AI workloads using the Ray distributed computing framework, with a focus on Torc
In good faith and with no judgment (mistakes happen), I truly hope that Anthropic will hear the feedback and change course on this. Anthropic is a company that has been raising awareness about AI mani
Automation startup Neura Robotics GmbH is raising a funding round worth up to 1.4 billion from a group of prominent investors. The company detailed today that the consortium includes Amazon.com Inc.,
arXiv:2503.13358v5 Announce Type: replace Abstract: Diffusion models for super-resolution (SR) produce high-quality visual results but require expensive computational costs. Despite the development of
DiffusionGemma is an experimental open model built for exceptionally fast text generation that NVIDIA has optimized to run on GeForce RTX GPUs, RTX PRO, and DGX Spark systems. Rather than generating t
arXiv:2605.27770v2 Announce Type: cross Abstract: We introduce `dualGNN', an autoregressive message-passing GNN for sampling fine, regular triangulations (FRTs) of convex polytopes. dualGNN operates o
Financial Times: Sources: Germany's Neura Robotics, which builds AI-powered humanoid robots, raised 1.4B from Tether, Qualcomm, Amazon, Nvidia, and others at a ~7B valuation — Crypto group Tether, Ama
Anissa Gardizy / The Information: Sources: OpenAI is in advanced talks to lease a proposed 10GW data center campus in Ohio as part of a deal that could include financial backing from Nvidia — OpenAI i
arXiv:2606.10445v1 Announce Type: cross Abstract: Semi-structured 2:4 sparsity is widely supported by modern accelerators, providing up to a 2x theoretical speedup. However, its strict 50% sparsity co
arXiv:2601.13994v3 Announce Type: replace-cross Abstract: Differentiable sparse linear algebra is foundational for scientific machine learning, yet PyTorch lacks a unified library for it: torch.sparse
arXiv:2606.10706v1 Announce Type: cross Abstract: Resource constraints increasingly determine what can be trained, fine-tuned, and deployed in large language models (LLMs), yet efficiency is often stu
Usage share of OpenAI grew vs Anthropic yesterday despite Mythos 5 / Fable 5 launch Multiple power users at SemiAnalysis tried Mythos / Fable Got refusals for nonsensical reasons Got pissed off at Ant
arXiv:2606.07574v1 Announce Type: cross Abstract: Manifold-constrained hyper-connections (mHCs) have recently been proposed as a principled extension of hyper-connections, where the residual mixing ma
NVIDIA FLARE Auto-FL automates federated learning research by constraining agent actions through a control plane, enforcing fixed benchmark contracts, and using an experiment ledger to ensure reproduc
arXiv:2606.08002v1 Announce Type: new Abstract: Underwater instance segmentation integrates pixel-level mask prediction and instance-level discrimination for marine resource exploration, ecological mo
arXiv:2602.14033v1 Announce Type: cross Abstract: Future sixth-generation (6G) mobile networks will demand artificial intelligence (AI) agents that are not only autonomous and efficient, but also capa
BREAKING NEWS: Anthropic's latest model will NOT help you if it thinks your ML research/ML engineering is interesting, and/or will secretly degrade its IQ so that the average engineer won't notice. We
This article from SemiAnalysis tracks the performance evolution of DeepSeek V4, a 1.6 trillion parameter model, over its first 43 days of training on Huawei and NVIDIA hardware including GB300 NVL72,
AI token economics is creating a data normalization crisis across the technology stack as enterprises scramble to apply consistent financial controls to GPU clusters, AI factories and token-based cons
for those keeping track at home it was 34 days between signing this deal and launching Mythos-class model GA to the world. https://x.com/leerob/status/2052059466821198061?s=20 building on @nvidia stac
arXiv:2606.09551v1 Announce Type: cross Abstract: Two-server secure inference allows a client to query a hosted large language model (LLM) without revealing prompts or embeddings. Recent GPU systems b
@GaryMarcus Hey Gary, I'm reading one of your books right now, ‘Taming Silicon Valley’. I have to say, it's really opened my eyes to what's going on out there in the AI industry! Looking forward to yo
GPU MODE has powered much of the public GPU kernel work online, with a permissive license from day one and generous credit from researchers, NVIDIA, AMD, and others. Today we’re moving our datasets to
arXiv:2606.07666v1 Announce Type: cross Abstract: Noisy intermediate-scale quantum (NISQ) processors are entering an early fault-tolerance regime where full quantum error correction carries prohibitiv
Nvidia Cosmos 3 is an open physical AI foundation model built on a mixture-of-transformers architecture that combines vision reasoning, world generation, and action prediction for reasoning, simulatio
arXiv:2606.09112v1 Announce Type: cross Abstract: The rapid evolution of artificial intelligence has led to substantial advances in deep neural networks. Nonetheless, conventional GPU-based training r
arXiv:2602.10016v3 Announce Type: replace-cross Abstract: Deriving predictable scaling laws that govern the relationship between model performance and computational investment is crucial for designing
arXiv:2601.20408v2 Announce Type: replace-cross Abstract: Enterprise LLM deployment faces a critical scalability challenge: organizations must optimize models systematically to scale AI initiatives wi
arXiv:2606.08039v1 Announce Type: new Abstract: Robotic simulators are a cornerstone of modern research in aerial robotics, serving both as a vehicle for the development of new control algorithms and
NVIDIA GPUs with Confidential Computing are now used for confidential inference in Apple’s Private Cloud Compute (PCC), as it expands beyond Apple’s data centers to Google Cloud. Unveiled during Apple
arXiv:2606.09200v1 Announce Type: cross Abstract: The rapid growth of large-scale machine learning (ML) has made distributed training across multiple GPUs a fundamental component of modern ML systems.
In this post, we show how to train robot policies for the Unitree H1 humanoid with NVIDIA Isaac Lab on Amazon SageMaker AI across two compute options: Amazon SageMaker HyperPod and Amazon SageMaker Tr
arXiv:2606.09679v1 Announce Type: new Abstract: We describe our system for the SoccerNet 2026 Player-Centric Ball-Action Spotting Challenge, which requires predicting who performs which action and whe
arXiv:2606.08382v1 Announce Type: cross Abstract: Low-rank projection has emerged as a promising approach for compressing the KV cache by exploiting hidden-dimension redundancy. However, prior methods
arXiv:2606.09516v1 Announce Type: new Abstract: Real-time video restoration (VR) for live streams requires high-resolution outputs under strict per-frame latency constraints. Existing one-step diffusi
arXiv:2510.16028v4 Announce Type: replace-cross Abstract: Neural networks increasingly run on hardware outside the user's control (cloud GPUs, inference marketplaces). Yet ML-as-a-Service reveals litt