<?xml version="1.0" encoding="UTF-8"?><urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:news="http://www.google.com/schemas/sitemap-news/0.9" xmlns:xhtml="http://www.w3.org/1999/xhtml" xmlns:image="http://www.google.com/schemas/sitemap-image/1.1" xmlns:video="http://www.google.com/schemas/sitemap-video/1.1"><url><loc>https://lizeman.github.io/llm-arch-kb/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/about/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/bigbird/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/dsa/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/flash-attention/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/gqa/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/lightning-attention/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/linear-attention/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/linformer/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/mha/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/mla/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/mqa/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/performer/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/reformer/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/sliding-window/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/attention/sparse-transformer/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/compare/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/facets/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/facets/efficiency/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/facets/hardware-aware/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/facets/inference-only/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/facets/kv-cache/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/facets/long-context/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/facets/parameter-free/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/facets/quality/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/facets/routing/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/facets/training-stability/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/aux-loss-free/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/deepseek-moe/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/ffn-relu/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/geglu/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/gelu/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/gshard/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/kimi-k2-moe/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/mixtral-moe/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/reglu/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/sparse-moe/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/swiglu/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/ffn-moe/switch-transformer/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/glossary/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/long-context/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/long-context/activation-beacon/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/long-context/compressive-transformer/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/long-context/landmark-attention/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/long-context/longnet/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/long-context/memorizing-transformers/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/long-context/streaming-llm/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/command-r-plus-104b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/deepseek-r1/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/deepseek-v1-67b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/deepseek-v2/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/deepseek-v3-1/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/deepseek-v3-2/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/deepseek-v3/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/deepseek-v4-pro/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/gemma-1-7b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/gemma-2-27b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/gemma-3-27b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/glm-4-5/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/glm-4-9b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/glm-130b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/gpt-4/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/gpt-oss-120b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/hunyuan-large-389b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/kimi-k1-5/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/kimi-k2/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/kimi-linear-48b-a3b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/llama-1-65b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/llama-2-70b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/llama-3-1-70b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/llama-3-3-70b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/llama-4-scout/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/minimax-01/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/minimax-m1/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/mistral-7b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/mistral-large-3-675b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/mixtral-8x7b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/nemotron-3-nano-30b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/olmo-1-7b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/olmo-2-13b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/olmo-3-32b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/olmoe-1b-7b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/phi-4-14b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/qwen-2-5-72b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/qwen3-30b-a3b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/qwen3-32b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/qwen3-235b-a22b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/qwen3-next-80b-a3b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/models/recurrent-gemma-2b/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/normalization/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/normalization/gemma3-norm-everywhere/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/normalization/layernorm/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/normalization/ngpt/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/normalization/norm-placement/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/normalization/olmo2-post-norm/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/normalization/qk-norm/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/normalization/rmsnorm/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/normalization/sandwich-ln/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/paths/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/paths/modern-attention/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/paths/modern-normalization/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/paths/moe-basics/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/positional/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/positional/alibi/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/positional/decoupled-rope/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/positional/longrope/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/positional/nope/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/positional/ntk-aware-rope/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/positional/position-interpolation/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/positional/rope/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/positional/sinusoidal/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/positional/yarn/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/references/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/residual/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/residual/deepnet/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/residual/dynamic-hc/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/residual/hyper-connections/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/residual/normformer/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/residual/residual-overview/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/residual/rezero/</loc><lastmod>2026-05-12T00:00:00.000Z</lastmod></url><url><loc>https://lizeman.github.io/llm-arch-kb/search/</loc></url><url><loc>https://lizeman.github.io/llm-arch-kb/timeline/</loc></url></urlset>