Stefanie Jegelka

Transformers learn translation, rotation, and scale invariance through distinct, identifiable mechanisms, paving the way for structured equivariant function learning.

Merging transformers sequentially along their execution path, rather than layer-by-layer, dramatically reduces error accumulation and unlocks principled multi-task capacity expansion.

Partial data augmentation can match the statistical benefits of full augmentation, challenging the notion that complete symmetry is necessary for optimal learning.