Flow Matching for Generative Modeling · HackerPapers discussions

Flow Matching for Generative Modeling — mentioned by ecrasevvv

Fri, 24 Jul 2026 09:42:53 GMT

Hi, the "Flow Matching paradigm" a simulation-free approach for training Continuous Normalizing Flows (CNFs) based on regressing vector fields of fixed conditional probability paths (original paper: https://arxiv.org/abs/2210.02747, MIT Course with practical examples: https://diffusion.csail.mit.edu/2026/index.html). Less formally Flow Matching can be described as "a technique to learn how to transport samples from one distribution to another". For example we could learn how to transport samples from a simple distribution we can easily sample from (e.g. Gaussian) to a complex distribution…

Paper and discussion history. A mention is not an endorsement.

Flow Matching for Generative Modeling — mentioned by jc4p

Fri, 27 Jun 2025 23:21:49 GMT

i've been trying to keep up with this field (image generation) so here's quick notes I took: Claude's Summary: "Normalizing flows aren't dead, they just needed modern techniques" My Summary: "Transformers aren't just for text" 1. SOTA model for likelihood on ImageNet 64×64, first ever sub 3.2 (Bits Per Dimension) prev was 2.99 by a hybrid diffusion model 2. Autoregressive (transformers) approach, right now diffusion is the most popular in this space (it's much faster but a diff approach) tl;dr of autoregressive vs diffusion (there's also other approaches) Autoregression: step based, generate…

Paper and discussion history. A mention is not an endorsement.

Flow Matching for Generative Modeling — mentioned by lnyan

Tue, 20 May 2025 19:46:35 GMT

Note that it's very likely that Veo models are based on "Flow Matching" [1] [1] https://arxiv.org/abs/2210.02747

Paper and discussion history. A mention is not an endorsement.