Lars W. Höfer

Latent Meta-Abstraction over Transformer Representations: A Two-System Architecture for Persistent Meta-Abstraction and Compute Leverage

This concept paper proposes a two-system architecture pairing a conventional transformer (Primary Model) with a separately parameterized Meta-Abstraction Model restricted operationally to activation-derived Primary representations and its own latent memory. A smaller or lower-cos...

Latent Meta-Abstraction over Transformer Representations: A Two-System Architecture for Persistent Meta-Abstraction and Compute Leverage

This concept paper proposes a two-system architecture pairing a conventional transformer (Primary Model) with a separately parameterized Meta-Abstraction Model restricted operationally to activation-derived Primary representations and its own latent memory. A smaller or lower-cos...