Cover Image

How Compositional Generalization and Creativity Improve as Diffusion Models are Trained

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Favero, Alessandro, Sclocchi, Antonio, Cagnetta, Francesco, Frossard, Pascal, Wyart, Matthieu
Format:	Preprint
Published:	2025
Subjects:	Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Scaling Laws and Representation Learning in Simple Hierarchical Languages: Transformers vs. Convolutional Architectures
by: Cagnetta, Francesco, et al.
Published: (2025)

Bigger Isn't Always Memorizing: Early Stopping Overparameterized Diffusion Models
by: Favero, Alessandro, et al.
Published: (2025)

A Phase Transition in Diffusion Models Reveals the Hierarchical Nature of Data
by: Sclocchi, Antonio, et al.
Published: (2024)

How Deep Neural Networks Learn Compositional Data: The Random Hierarchy Model
by: Cagnetta, Francesco, et al.
Published: (2023)

Probing the Latent Hierarchical Structure of Data via Diffusion Models
by: Sclocchi, Antonio, et al.
Published: (2024)

Towards a theory of how the structure of language is acquired by deep neural networks
by: Cagnetta, Francesco, et al.
Published: (2024)

On the different regimes of Stochastic Gradient Descent
by: Sclocchi, Antonio, et al.
Published: (2023)

Learning curves theory for hierarchically compositional data with power-law distributed features
by: Cagnetta, Francesco, et al.
Published: (2025)

Deep networks learn to parse uniform-depth context-free languages from local statistics
by: Parley, Jack T., et al.
Published: (2026)

Deriving Neural Scaling Laws from the statistics of natural language
by: Cagnetta, Francesco, et al.
Published: (2026)

Task Addition and Weight Disentanglement in Closed-Vocabulary Models
by: Hazimeh, Adam, et al.
Published: (2025)

Learn from your own latents and not from tokens: A sample-complexity theory
by: Korchinski, Daniel J., et al.
Published: (2026)

Sparse Training of Discrete Diffusion Models for Graph Generation
by: Qin, Yiming, et al.
Published: (2023)

How Deep Networks Learn Sparse and Hierarchical Data: the Sparse Random Hierarchy Model
by: Tomasini, Umberto, et al.
Published: (2024)

Backdoor Unlearning by Linear Task Decomposition
by: Abdelraheem, Amel, et al.
Published: (2025)

MEMOIR: Lifelong Model Editing with Minimal Overwrite and Informed Retention for LLMs
by: Wang, Ke, et al.
Published: (2025)

Hierarchical Concept Geometry in Language Models Emerges from Word Co-occurrence
by: Nava, Andres, et al.
Published: (2026)

Graph-Dictionary Signal Model for Sparse Representations of Multivariate Data
by: Cappelletti, William, et al.
Published: (2024)

LiNeS: Post-training Layer Scaling Prevents Forgetting and Enhances Model Merging
by: Wang, Ke, et al.
Published: (2024)

Generative Modelling of Structurally Constrained Graphs
by: Madeira, Manuel, et al.
Published: (2024)

PUMA: margin-based data pruning
by: Maroto, Javier, et al.
Published: (2024)

The Physics of Data and Tasks: Theories of Locality and Compositionality in Deep Learning
by: Favero, Alessandro
Published: (2025)

Flow based approach for Dynamic Temporal Causal models with non-Gaussian or Heteroscedastic Noises
by: Rahmani, Abdellah, et al.
Published: (2025)

Deep End-to-End Survival Analysis with Temporal Consistency
by: Vieyra, Mariana Vargas, et al.
Published: (2024)

Causal Temporal Regime Structure Learning
by: Rahmani, Abdellah, et al.
Published: (2023)

DeFoG: Discrete Flow Matching for Graph Generation
by: Qin, Yiming, et al.
Published: (2024)

Model soups need only one ingredient
by: Abdollahpoorrostam, Alireza, et al.
Published: (2026)

Pareto Low-Rank Adapters: Efficient Multi-Task Learning with Preferences
by: Dimitriadis, Nikolaos, et al.
Published: (2024)

Bures-Wasserstein Means of Graphs
by: Haasler, Isabel, et al.
Published: (2023)

Localizing Task Information for Improved Model Merging and Compression
by: Wang, Ke, et al.
Published: (2024)

Sampling Data with Chains of Forward-Backward Diffusion Steps
by: Kang, Hyunmo, et al.
Published: (2026)

Generating Directed Graphs with Dual Attention and Asymmetric Encoding
by: Carballo-Castro, Alba, et al.
Published: (2025)

Does Generation Require Memorization? Creative Diffusion Models using Ambient Diffusion
by: Shah, Kulin, et al.
Published: (2025)

Balancing Symmetry and Efficiency in Graph Flow Matching
by: Honoré, Benjamin, et al.
Published: (2026)

Training Data Protection with Compositional Diffusion Models
by: Golatkar, Aditya, et al.
Published: (2023)

On the Emergence of Linear Analogies in Word Embeddings
by: Korchinski, Daniel J., et al.
Published: (2025)

How Creative Are Large Language Models in Generating Molecules?
by: Tao, Wen, et al.
Published: (2026)

rETF-semiSL: Semi-Supervised Learning for Neural Collapse in Temporal Data
by: Xie, Yuhan, et al.
Published: (2025)

Inductive Domain Transfer In Misspecified Simulation-Based Inference
by: Senouf, Ortal, et al.
Published: (2025)

Logical Guidance for the Exact Composition of Diffusion Models
by: Alesiani, Francesco, et al.
Published: (2026)