Hyper-Transforming Latent Diffusion Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Peis, Ignacio, Koyuncu, Batuhan, Valera, Isabel, Frellsen, Jes
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908623643344896
author Peis, Ignacio
Koyuncu, Batuhan
Valera, Isabel
Frellsen, Jes
author_facet Peis, Ignacio
Koyuncu, Batuhan
Valera, Isabel
Frellsen, Jes
contents We introduce a novel generative framework for functions by integrating Implicit Neural Representations (INRs) and Transformer-based hypernetworks into latent variable models. Unlike prior approaches that rely on MLP-based hypernetworks with scalability limitations, our method employs a Transformer-based decoder to generate INR parameters from latent variables, addressing both representation capacity and computational efficiency. Our framework extends latent diffusion models (LDMs) to INR generation by replacing standard decoders with a Transformer-based hypernetwork, which can be trained either from scratch or via hyper-transforming: a strategy that fine-tunes only the decoder while freezing the pre-trained latent space. This enables efficient adaptation of existing generative models to INR-based representations without requiring full retraining. We validate our approach across multiple modalities, demonstrating improved scalability, expressiveness, and generalization over existing INR-based generative models. Our findings establish a unified and flexible framework for learning structured function representations.
format Preprint
id arxiv_https___arxiv_org_abs_2504_16580
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Hyper-Transforming Latent Diffusion Models
Peis, Ignacio
Koyuncu, Batuhan
Valera, Isabel
Frellsen, Jes
Machine Learning
We introduce a novel generative framework for functions by integrating Implicit Neural Representations (INRs) and Transformer-based hypernetworks into latent variable models. Unlike prior approaches that rely on MLP-based hypernetworks with scalability limitations, our method employs a Transformer-based decoder to generate INR parameters from latent variables, addressing both representation capacity and computational efficiency. Our framework extends latent diffusion models (LDMs) to INR generation by replacing standard decoders with a Transformer-based hypernetwork, which can be trained either from scratch or via hyper-transforming: a strategy that fine-tunes only the decoder while freezing the pre-trained latent space. This enables efficient adaptation of existing generative models to INR-based representations without requiring full retraining. We validate our approach across multiple modalities, demonstrating improved scalability, expressiveness, and generalization over existing INR-based generative models. Our findings establish a unified and flexible framework for learning structured function representations.
title Hyper-Transforming Latent Diffusion Models
topic Machine Learning
url https://arxiv.org/abs/2504.16580