The Impact of Model Zoo Size and Composition on Weight Space Learning

Damian Falk, Konstantin Schürholt, Damian Borth · 2025

Read The Impact of Model Zoo Size and Composition on Weight Space Learning by Damian Falk, Konstantin Schürholt, Damian Borth (2025) 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-2504-10141.txt.

Description

Damian Falk and co-authors study weight space learning, a field that treats the trained weights of neural networks as data. Teams reuse saved models to cut training cost, but most past work needs all models to share one exact design. Their 2025 paper lifts that limit.

The authors adapt a known method so it can train on a mixed set of models with unlike designs. They then test how the size and mix of that set change the results. Adding models trained on varied image sets had a large effect on skill and on transfer to new tasks. Code for the work is posted in a public code archive.

Catalog record

CREATORDamian Falk, Konstantin Schürholt, Damian Borth
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-10141

Browse the full library · About the Tower of Babel