テクノロジー
GoogleとIntel、AI基盤の協業を深める：XeonとIPUでGoogle Cloudの土台を再強化
GoogleとIntelは2026年4月9日、AIとクラウド基盤を巡る協業を深めると発表した。今回の発表の軸は、派手な新型AIアクセラレータの投入ではない。Googleの大規模インフラで複数世代のIntel Xeonを継 […]
Term
C4
別名: Colossal Clean Crawled Corpus, C4
Overview
最終更新: 2026年7月11日The Colossal Clean Crawled Corpus (C4) is a massive, cleaned version of the Common Crawl web crawl corpus. It was originally introduced by Google for training the T5 model and has since become a standard dataset for pre-training various large language models, including versions of GPT.
Mentioned Articles
2 件
External Mentions
6 件- arXiv Joint Structural Pruning and Mixed-Precision Quantization for LLM Compression
- arXiv Towards the Connection between Activation Sparsity and Flat Minima
- arXiv Maximum Lifetime of the Vegetative Biosphere
- arXiv The Economics of Model Collapse: Equilibrium, Welfare, and Optimal Provenance Subsidies in Synthetic Data Markets
- arXiv PyNMC: An Open-Source Framework for Neutron Multiplicity Counting Simulation Coupling OpenMC, FREYA, and ALPHANSO
- Hacker News Millions of Tiny Databases [pdf]