Saved in:
Bibliographic Details
Main Authors: Zhmoginov, Andrey, Lee, Jihwan, Sandler, Mark
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.05641
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909640916205568
author Zhmoginov, Andrey
Lee, Jihwan
Sandler, Mark
author_facet Zhmoginov, Andrey
Lee, Jihwan
Sandler, Mark
contents Modern Foundation Models (FMs) are typically trained on corpora spanning a wide range of different data modalities, topics and downstream tasks. Utilizing these models can be very computationally expensive and is out of reach for most consumer devices. Furthermore, most of the broad FM knowledge may actually be irrelevant for a specific task at hand. Here we explore a technique for mapping parameters of a large Transformer to parameters of a smaller specialized model. By making this transformation task-specific, we aim to capture a narrower scope of the knowledge needed for performing a specific task by a smaller model. We study our method on image modeling tasks, showing that performance of generated models exceeds that of universal conditional models.
format Preprint
id arxiv_https___arxiv_org_abs_2506_05641
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Projectable Models: One-Shot Generation of Small Specialized Transformers from Large Ones
Zhmoginov, Andrey
Lee, Jihwan
Sandler, Mark
Machine Learning
Computation and Language
Modern Foundation Models (FMs) are typically trained on corpora spanning a wide range of different data modalities, topics and downstream tasks. Utilizing these models can be very computationally expensive and is out of reach for most consumer devices. Furthermore, most of the broad FM knowledge may actually be irrelevant for a specific task at hand. Here we explore a technique for mapping parameters of a large Transformer to parameters of a smaller specialized model. By making this transformation task-specific, we aim to capture a narrower scope of the knowledge needed for performing a specific task by a smaller model. We study our method on image modeling tasks, showing that performance of generated models exceeds that of universal conditional models.
title Projectable Models: One-Shot Generation of Small Specialized Transformers from Large Ones
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2506.05641