InFandhebyNancy·Sep 9What Is Flow Matching? 11 Years From NICE to FLUX.2The lineage from NICE (2014) to Stable Diffusion 3 and FLUX.2, and why a straight training path is not a straight generation trajectory.
InTowards AIbyPhrugsa Limbunlom (Gift)·Aug 27Generative Modeling with Flow Matching, Optimal Transport, and Schrödinger bridgeA unified formulation of flow matching, optimal transport, and the Schrödinger bridge as transport between distributions.A response icon3A response icon3
Steins·Aug 262 Reasons Why Stable Diffusion 3 is NOT "Diffusion"!Stable Diffusion 3 Clearly Explained!
shashank Jain·Jul 23Teaching noise to become a picture: flow matching from scratchThis is a walkthrough of a small project I ran on my laptop. I trained a flow matching model on MNIST digits, then on CelebA faces, and…
Shubham Sahoo·Jun 11DDPM vs Flow Matching on the Same Architecture: What the Papers Don’t Tell You About Training…Building DDPM, DDIM, and Flow Matching on a free Colab GPU — the bugs, the numbers, and what the papers skip.A response icon1A response icon1
Siamak Yousefi·May 24Overview of Reinforcement Learning Methods for Enhancing VLA Robot TrainingThe Vision Language Action (VLA) Models have gained tremendous attention in the past couple of years, thanks to the great progress made in…
InToward HumanoidsbyYuri Fung·May 5What actually improved Push-T?Flow Matching is hard to beatA response icon1A response icon1
InToward HumanoidsbyYiChun Liao·May 2One-step Generative Modeling with Flow matching, Mean Flows and Solution Flow Models for Robotic…A Comparison on the LIBERO dataset
InTowards AIbyPhrugsa Limbunlom (Gift)·Apr 20Physics-Inspired Generative Modeling: Diffusion, Flow Matching, and Energy-Based ModelsFrom Thermodynamics to Neural Networks: A Unified Framework for Modern Generative AIA response icon1A response icon1
InToward HumanoidsbyAntony Zhao·Apr 20Flow Matching Transformer For PushTTraining Flow Matching to Mimic Expert Demonstrations
