Pedro Domingos: It's Time to Call Residual Streams What They Are — RAM

pmddomingos · x · 2026-09-10

Prominent ML researcher and author Pedro Domingos argued that residual streams in Transformers should simply be called RAM, pointing out they act as the model's memory that layers read from and write to. The analogy is sparking discussion about how we name and understand Transformer internals.

Original post →

More from Models

Models channel →