https://puyang.me/blog/engineering_system_view/training-models-larger-than-one-gpu-en.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/engineering_system_view/training-models-larger-than-one-gpu.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/engineering_system_view/what-are-we-optimizing-in-agent-context-en.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/engineering_system_view/what-are-we-optimizing-in-agent-context.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/geometry-of-transformers/what-attention-does-en.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/geometry-of-transformers/what-attention-does.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/geometry-of-transformers/why-multi-head-matters-en.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/geometry-of-transformers/why-multi-head-matters.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/high-dimensional-space-and-machine-learning/distance-breakdown-en.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/high-dimensional-space-and-machine-learning/distance-breakdown.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/high-dimensional-space-and-machine-learning/hypersphere-en.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/high-dimensional-space-and-machine-learning/hypersphere.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/high-dimensional-space-and-machine-learning/orthogonality-en.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/high-dimensional-space-and-machine-learning/orthogonality.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/https://puyang.me/blog/representation-space-of-large-models/semantic-linearity-en.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/representation-space-of-large-models/semantic-linearity.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/representation-space-of-large-models/spherical-coding-en.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/representation-space-of-large-models/spherical-coding.html2026-03-11T21:43:36.000Zhttps://puyang.me/blog/theory-of-tokenizers/what-tokenization-does-en.html2026-03-12T21:32:20.000Zhttps://puyang.me/blog/theory-of-tokenizers/what-tokenization-does.html2026-03-12T21:32:20.000Zhttps://puyang.me/blog/theory-of-tokenizers/why-character-level-rarely-wins-en.html2026-03-12T21:23:04.000Zhttps://puyang.me/blog/theory-of-tokenizers/why-character-level-rarely-wins.html2026-03-12T21:23:04.000Zhttps://puyang.me/blog/theory-of-tokenizers/why-vocab-size-stays-near-50k-en.html2026-03-12T21:23:04.000Zhttps://puyang.me/blog/theory-of-tokenizers/why-vocab-size-stays-near-50k.html2026-03-12T21:23:04.000Zhttps://puyang.me/2026-03-11T00:20:29.000Zhttps://puyang.me/news/2025-12-05-paper-accepted.html2026-03-09T21:21:27.000Zhttps://puyang.me/news/2026-01-20-launched-new-site.html2026-03-09T21:21:27.000Zhttps://puyang.me/news/2026-03-09T21:21:27.000Zhttps://puyang.me/projects/2026-03-20T19:21:10.000Zhttps://puyang.me/projects/network-architecture-for-6g-network/2026-03-20T19:21:10.000Zhttps://puyang.me/projects/sample-project/2026-03-20T19:21:10.000Zhttps://puyang.me/projects/semantic-routing-and-load-balance/2026-03-20T19:21:10.000Zhttps://puyang.me/projects/wireless-avionics-intra-communications-waic-system/2026-03-20T19:21:10.000Zhttps://puyang.me/publications/