ruvector-attn-mincut vs topological-coherence vs ruvector-attention-wasm

Min-cut gating attention operator: dynamic graph-based alternative to softmax attention

Toroidal topology primitives for LLM coherence research — Tonnetz geometry, spectral gap, attention masks, Karmonic spectral filter (v7: inference-time bias null, training-time valid)

High-performance WebAssembly attention mechanisms: Multi-Head, Flash, Hyperbolic, MoE, CGT Sheaf Attention with GPU acceleration for transformers and LLMs