AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

PairFlow: Closed-Form Source-Target Coupling for Few-Step Generation in Discrete Flow Models

Mingue Park, Jisung Hwang, Seungwoo Yoo, Kyeongmin Yeo, Minhyuk Sung

ICLR 2026 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Flow matching and flow sampling · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train UDLM on pairs from closed-form backward inversion (PairFlow)UDLM · QM9 1-stepvalid molecules: 17.5 -> 223.4Improved
use closed-form paired data instead of random uniform source pairing (ablation)UDLM · QM9 1-stepvalid molecules: 47.1 -> 223.4Improved
use closed-form paired data instead of random uniform source pairing (ablation)UDLM · QM9 64-stepvalid molecules: 998.6 -> 1000.2No clear effect
PairFlow preprocessing instead of ReDi finetuningUDLM · QM9 2-stepvalid molecules: 232.4 -> 416.0Improved
PairFlow preprocessing instead of DCD finetuningUDLM · QM9 2-stepvalid molecules: 530.8 -> 416.0Hurt
distill from PairFlow base instead of UDLM base with DCDUDLM + DCD · QM9 2-stepvalid molecules: 530.8 -> 685.8Improved
distill from PairFlow base with ReDiUDLM + ReDi · ZINC-250k 2-stepvalid molecules: 75.9 -> 221.5Improved
train with PairFlow pairsUDLM · MNIST-Binary 1-stepFID: about 40.59 vs UDLM; 68.9% improvementImproved
PairFlow instead of ReDi at 2 stepsUDLM · MNIST-Binary 2-stepFID: 10.36 -> 15.61Hurt
train with PairFlow pairsUDLM · CIFAR-10FID: 33.65 -> 28.07Improved
apply DCD or ReDi distillationUDLM · CIFAR-10FID / ISHurt
build pairs with backward inversion rather than forward-velocity sampling (ablation)PairFlow pair construction · training set coveragecoverage: forward sampling covers about 63% of training setImproved

Later papers that compared with this method

1 later accepted paper reports PairFlow as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy