Princeton reframes the residual stream as a geometric flow

Claim

Princeton introduced Deep Delta Learning, treating the transformer residual stream as a continuous flow that cleans its own feature subspaces layer by layer.

Reporting date: 2026-01-03 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

• Welcome to January 3, 2026 (theinnermostloop.substack.com)