few-step-generation

Oct 03, 2026

LIT-775: Inductive Moment Matching

Oct 03, 2026

LIT-783: Large Scale Diffusion Distillation via Score-Regularized Continuous-Time Consistency

Oct 03, 2026

LIT-784: Mean Flows for One-step Generative Modeling

Oct 03, 2026

LIT-786: Improved Techniques for Training Consistency Models

Oct 03, 2026

LIT-787: One Step Diffusion via Shortcut Models

Oct 03, 2026

LIT-790: Simplifying, Stabilizing and Scaling Continuous-Time Consistency Models

Oct 03, 2026

LIT-791: Improving the Training of Rectified Flows

Oct 03, 2026

LIT-793: Flow map matching with stochastic interpolants: A mathematical framework for consistency models

Oct 03, 2026

LIT-796: InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generation

Oct 03, 2026

SOTA-443: Down-weight large residuals in consistency and MeanFlow training: use a Pseudo-Huber loss or an adaptive inverse-error weight, not plain squared L2

Oct 03, 2026

SOTA-446: In consistency training, take the target from the current weights under stop-gradient, not from an EMA teacher

Oct 03, 2026

SOTA-447: When distilling a large image or video model by continuous-time consistency, add a small DMD distribution-matching term

Oct 03, 2026

SOTA-448: Without a pretrained teacher, train a one-step generator as a MeanFlow average-velocity model rather than by consistency training or a shortcut model

Oct 03, 2026

THEORY-118: Continuous-time consistency training is unstable through the time derivative in its tangent, and tangent normalization, which MeanFlow's adaptive loss weight already is, controls it

Oct 03, 2026

THEORY-119: Consistency, shortcut, MeanFlow and progressive-distillation models all estimate the two-time flow map of the probability-flow ODE, and differ in which identity of that map they train on

Oct 03, 2026

THEORY-121: Distribution-matching distillation loses coverage because it descends a reverse divergence on the student's own samples, and a consistency term on teacher trajectories restores it

Oct 03, 2026

THEORY-123: A flow's sampling ODE curves because the noise-data pairs it was trained on cross and the learned velocity averages them, so a coupling whose straight paths do not cross gives a straight flow

Oct 03, 2026

THEORY-124: Consistency training is a bootstrapped fixed-point iteration, not descent on a loss, so a target that lags the student swamps a data signal that enters only at the order of the step

Oct 02, 2026

LIT-770: DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation

Sep 24, 2026

LIT-629: Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion

Sep 24, 2026

LIT-631: From Slow Bidirectional to Fast Autoregressive Video Diffusion Models

Sep 24, 2026

LIT-643: One-step Diffusion with Distribution Matching Distillation

Sep 24, 2026

LIT-646: Improved Distribution Matching Distillation for Fast Image Synthesis

Sep 24, 2026

SOTA-392: To get a one-step sampler from a rectified flow, distil it; treat reflow as an optional extra pass, and keep the pre-reflow model for many steps

Sep 24, 2026

SOTA-394: When distilling a causal few-step video generator, take the teacher bidirectional, not causal

Sep 21, 2026

SOTA-289: When the sampling budget is small, prefer uniform-state discrete diffusion with consistency distillation — masked diffusion cannot revise what it has already emitted

Sep 10, 2026

SOTA-206: Keep a multi-step sampling option in a few-step generative model

Aug 24, 2026

LIT-067: Progressive Distillation for Fast Sampling of Diffusion Models

Aug 24, 2026

LIT-076: DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps

Aug 24, 2026

anthology-of-the-sota

Explorer

few-step-generation

Graph View

LIT-775: Inductive Moment Matching

LIT-783: Large Scale Diffusion Distillation via Score-Regularized Continuous-Time Consistency

LIT-784: Mean Flows for One-step Generative Modeling

LIT-786: Improved Techniques for Training Consistency Models

LIT-787: One Step Diffusion via Shortcut Models

LIT-790: Simplifying, Stabilizing and Scaling Continuous-Time Consistency Models

LIT-791: Improving the Training of Rectified Flows

LIT-793: Flow map matching with stochastic interpolants: A mathematical framework for consistency models

LIT-796: InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generation

SOTA-443: Down-weight large residuals in consistency and MeanFlow training: use a Pseudo-Huber loss or an adaptive inverse-error weight, not plain squared L2

SOTA-446: In consistency training, take the target from the current weights under stop-gradient, not from an EMA teacher

SOTA-447: When distilling a large image or video model by continuous-time consistency, add a small DMD distribution-matching term

SOTA-448: Without a pretrained teacher, train a one-step generator as a MeanFlow average-velocity model rather than by consistency training or a shortcut model

THEORY-118: Continuous-time consistency training is unstable through the time derivative in its tangent, and tangent normalization, which MeanFlow's adaptive loss weight already is, controls it

THEORY-119: Consistency, shortcut, MeanFlow and progressive-distillation models all estimate the two-time flow map of the probability-flow ODE, and differ in which identity of that map they train on

THEORY-121: Distribution-matching distillation loses coverage because it descends a reverse divergence on the student's own samples, and a consistency term on teacher trajectories restores it

THEORY-123: A flow's sampling ODE curves because the noise-data pairs it was trained on cross and the learned velocity averages them, so a coupling whose straight paths do not cross gives a straight flow

THEORY-124: Consistency training is a bootstrapped fixed-point iteration, not descent on a loss, so a target that lags the student swamps a data signal that enters only at the order of the step

LIT-770: DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation

LIT-629: Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion

LIT-631: From Slow Bidirectional to Fast Autoregressive Video Diffusion Models

LIT-643: One-step Diffusion with Distribution Matching Distillation

LIT-646: Improved Distribution Matching Distillation for Fast Image Synthesis

SOTA-392: To get a one-step sampler from a rectified flow, distil it; treat reflow as an optional extra pass, and keep the pre-reflow model for many steps

SOTA-394: When distilling a causal few-step video generator, take the teacher bidirectional, not causal

SOTA-289: When the sampling budget is small, prefer uniform-state discrete diffusion with consistency distillation — masked diffusion cannot revise what it has already emitted

SOTA-206: Keep a multi-step sampling option in a few-step generative model

LIT-067: Progressive Distillation for Fast Sampling of Diffusion Models

LIT-076: DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps

LIT-093: Consistency Models