finding
active
finding:with-5-repeated-data-point-transcoder-learns-a-datapoint-feature-and-rotated-constituent-features-that-do-not-activate-on-p

With 5% repeated data point, transcoder learns a datapoint feature and rotated constituent features that do not activate on p

Demonstrates unfaithfulness in a simple setup.

Source paper

extracted_from
A Toy Model of Mechanistic (Un)Faithfulness
Chris Olah

Neighborhood — ranked by edge-count