arXiv:2504.16580v4 Announce Type: replace 
Abstract: 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.

تم تقديم إطار توليدي جديد يجمع بين التمثيلات العصبية الضمنية والشبكات الفائقة المستندة إلى المحولات، مما يعزز قدرات نماذج المتغيرات الكامنة. تتجاوز هذه الطريقة المبتكرة قيود الأساليب السابقة، مما يوفر قدرة تمثيل أفضل وكفاءة حسابية.

Se ha introducido un nuevo marco generativo que combina Representaciones Neurales Implícitas con hiperredes basadas en Transformadores, mejorando las capacidades de los modelos de variables latentes. Este enfoque innovador supera las limitaciones de los métodos anteriores, ofreciendo una mejor capacidad de representación y eficiencia computacional.

Un nouveau cadre génératif a été introduit, combinant les Représentations Neurales Implicites avec des hyperréseaux basés sur des Transformateurs, améliorant ainsi les capacités des modèles de variables latentes. Cette approche innovante surmonte les limitations des méthodes précédentes, offrant une meilleure capacité de représentation et une efficacité computationnelle.

A new generative framework has been introduced that combines Implicit Neural Representations with Transformer-based hypernetworks, enhancing the capabilities of latent variable models. This innovative approach overcomes the limitations of previous methods, offering improved representation capacity and computational efficiency.

Hyper-Transforming Latent Diffusion Models

One More Thing in AI – Your Shortcut to AI Mastery

Hyper-Transforming Latent Diffusion Models

Was this article worth reading? Share it

One More Thing in AI

LucidQuery AI

Humanize AI

Airparser

GPTHumanizer

The Visualizer

Ready to build your own newsroom?