Simplicity Bias in Transformers and their Ability to Learn Sparse Boolean Functions.

Satwik Bhattamishra, Arkil Patel, Varun Kanade, Phil Blunsom

Browse the full ACL paper archive.