HyperTransport: Amortized Conditioning of T2I Generative Models

Valentino Maiorca, Eleonora Gualdoni, Xavier Suau, Marco Cuturi, Luca Zappella, Pau Rodríguez · 2026

Read HyperTransport: Amortized Conditioning of T2I Generative Models by Valentino Maiorca, Eleonora Gualdoni, Xavier Suau, Marco Cuturi, Luca Zappella, Pau Rodríguez (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-08254.txt.

Description

Valentino Maiorca, Eleonora Gualdoni, Xavier Suau, Marco Cuturi, and coauthors make activation steering cheap for text-to-image models. Steering beats prompting for stable control. But each new concept needs its own optimization run on the target model. That takes minutes per concept. It fails when concept sets are large or change all the time.

HyperTransport trains a hypernetwork to do the mapping at once. It maps an embedding from a pretrained encoder to the intervention parameters for that concept. CLIP is the encoder in their setup. The one-time training removes the per-concept cost. New concepts can be steered as fast as they can be embedded. Control stays predictable across requests.

Catalog record

CREATORValentino Maiorca, Eleonora Gualdoni, Xavier Suau, Marco Cuturi, Luca Zappella, Pau Rodríguez
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE80 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.08254

Browse the full library · About the Tower of Babel