Attention explorer

Select a token and inspect the weights it assigns to other tokens. Reveal the causal boundary.

How this experiment works

Compute dot products between a selected query and the token vectors, divide by the square root of their dimension, apply a causal mask and softmax.

What it does not prove

The vectors are small, hand-authored teaching values. Attention weight is not an explanation of a model’s reasoning.

Try this challenge

Select the third token. Toggle the causal mask and see which future tokens become visible.

Related Trust Agent courses · Sources and licences

Enable JavaScript to run the experiment on your device.