← Back to Issue 7

Attention Visualizer

Every word looks at every other word. See how.

Interactive companion to Issue 7: Attention Is All You Need

Try an example:

Attention Heads

Attention Heatmap

0.0

1.0

Attention Beams

Click a token above or a heatmap row to see where it looks

Attention(Q, K, V) = softmax( Q · K^T / √d_k ) · V

Click a token to see the Q, K, V breakdown for that word.

Note: This visualizer uses heuristic-based attention simulation, not a real transformer model. The patterns shown illustrate the types of attention that different heads learn — positional proximity, syntactic relationships, semantic similarity, and global context. Real attention weights are learned from data, but the principles are the same.