docs(triattention-v3): add Section 4.8 boundary skip negative result

docs(triattention-v3): add Methodology section and renumber

docs: add TriAttention V3 paper

docs: add Apple Silicon callout pointing to ekryski/mlx-swift-lm

Remove nihilistau reference from asymmetric K/V paper

docs: update MLX speed numbers — 97-100% baseline with delegated KVCache

docs: add MM-NIAH benchmark results + mlx-vlm Quick Start

docs: compact mode uses C++ sdpa_vector_qv Metal kernel (precision bug fixed)

docs: update MLX results with verified compact mode benchmarks

docs: fix MLX compact description — uses mx.quantized_matmul not C++ kernel

docs: update MLX section with compact mode + C++ Metal kernel

Revert "docs: add compact mode to MLX section"

docs: add compact mode to MLX section

docs: fix MLX section — lazy compression, remove auto-compress claims

docs: update MLX results with accurate K+V benchmark data