← Back to Labs

LLM Attention Weights

Step through tokenisation, attention scoring, value mixing, and KV-cache reuse on one short prompt.

Context WindowSofiaopenedthe archivebecause she needed the log55%20%15%10%
Step 1 / 4Tokenisation breaks text into model-sized pieces

The model does not read whole ideas or words. It reads token pieces, and even this short sentence becomes a small list of learned IDs.

Arrow keys to navigate · R to reset

Tap dots to jump to any step

Read the full article →Take the quiz →