The letter

One letter a week, in your inbox.

The signal of the week, what shipped, what to try, and the editor's note. No tracking, no ads, nothing else.

We keep your address, your language and the date you joined, nothing else. Every letter has a one-click unsubscribe link that deletes the record.

← Accept All   Archive
Transformer CircuitsResearch

Tracing Attention Computation Through Feature Interactions

July 31

We describe and apply a method to explain attention patterns in terms of feature interactions, and integrate this information into attribution graphs.

paper
Read at Transformer Circuits ↗

More from Transformer Circuits on Accept All.