Generative flows on discrete state-spaces: Enabling multimodal flows with applications to protein co-design

A Campbell, J Yim, R Barzilay, T Rainforth… - arxiv preprint arxiv …, 2024‏ - arxiv.org
Combining discrete and continuous data is an important capability for generative models.
We present Discrete Flow Models (DFMs), a new flow-based model of discrete data that …

View-decoupled transformer for person re-identification under aerial-ground camera network

Q Zhang, L Wang, VM Patel… - Proceedings of the …, 2024‏ - openaccess.thecvf.com
Existing person re-identification methods have achieved remarkable advances in
appearance-based identity association across homogeneous cameras such as ground …

Tackling the singularities at the endpoints of time intervals in diffusion models

P Zhang, H Yin, C Li, X **e - Proceedings of the IEEE/CVF …, 2024‏ - openaccess.thecvf.com
Most diffusion models assume that the reverse process adheres to a Gaussian distribution.
However this approximation has not been rigorously validated especially at singularities …

Energy-Based Modelling for Discrete and Mixed Data via Heat Equations on Structured Spaces

T Schröder, Z Ou, Y Li… - Advances in Neural …, 2025‏ - proceedings.neurips.cc
Energy-based models (EBMs) offer a flexible framework for probabilistic modelling across
various data domains. However, training EBMs on data in discrete or mixed state spaces …

Discrete Modeling via Boundary Conditional Diffusion Processes

Y Gu, X Feng, L Huang, Y Wu, Z Zhou, W Zhong… - arxiv preprint arxiv …, 2024‏ - arxiv.org
We present an novel framework for efficiently and effectively extending the powerful
continuous diffusion processes to discrete modeling. Previous approaches have suffered …

Solving Prior Distribution Mismatch in Diffusion Models via Optimal Transport

Z Wang, S Li, C Wang, S Cao, N Lei, Z Luo - arxiv preprint arxiv …, 2024‏ - arxiv.org
In recent years, the knowledge surrounding diffusion models (DMs) has grown significantly,
though several theoretical gaps remain. Particularly noteworthy is prior error, defined as the …

Exploring Straighter Trajectories of Flow Matching with Diffusion Guidance

S **ng, J Cao, H Huang, XY Zhang, R He - arxiv preprint arxiv …, 2023‏ - arxiv.org
Flow matching as a paradigm of generative model achieves notable success across various
domains. However, existing methods use either multi-round training or knowledge within …

Elucidating Flow Matching ODE Dynamics with respect to Data Geometries

G Mishne, Z Wan, Q Wang, Y Wang - arxiv preprint arxiv:2412.18730, 2024‏ - arxiv.org
Diffusion-based generative models have become the standard for image generation. ODE-
based samplers and flow matching models improve efficiency, in comparison to diffusion …

Gradient-Free Analytical Fisher Information of Diffused Distributions

F Wang, H Yin, H Zhu, S Zhuang, C Zhang, H Zhao‏ - openreview.net
Diffusion models (DMs) have demonstrated powerful distributional modeling capabilities by
matching the first-order score of diffused distributions. Recent advancements have explored …