A small-multiples grid of causal attention maps — one miniature token-by-token heat matrix per (layer, head), sortable by row entropy or previous-token mass, with the picked head enlarged beside it under token labels on both axes and its attention mass split into four disjoint shares that add to exactly 100.
npx shadcn@latest add @zyeon/chart-attention-headsSign in or get a license at Zyeon UI to see its source and install it.