Pith. sign in

REVIEW 19 cited by

Improved Distribution Matching Distillation for Fast Image Synthesis

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2405.14867 v2 pith:VIZGSVQO submitted 2024-05-23 cs.CV

Improved Distribution Matching Distillation for Fast Image Synthesis

classification cs.CV
keywords teacherdistributiontrainingdistillationgeneratedlossone-stepquality
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Recent approaches have shown promises distilling diffusion models into efficient one-step generators. Among them, Distribution Matching Distillation (DMD) produces one-step generators that match their teacher in distribution, without enforcing a one-to-one correspondence with the sampling trajectories of their teachers. However, to ensure stable training, DMD requires an additional regression loss computed using a large set of noise-image pairs generated by the teacher with many steps of a deterministic sampler. This is costly for large-scale text-to-image synthesis and limits the student's quality, tying it too closely to the teacher's original sampling paths. We introduce DMD2, a set of techniques that lift this limitation and improve DMD training. First, we eliminate the regression loss and the need for expensive dataset construction. We show that the resulting instability is due to the fake critic not estimating the distribution of generated samples accurately and propose a two time-scale update rule as a remedy. Second, we integrate a GAN loss into the distillation procedure, discriminating between generated samples and real images. This lets us train the student model on real data, mitigating the imperfect real score estimation from the teacher model, and enhancing quality. Lastly, we modify the training procedure to enable multi-step sampling. We identify and address the training-inference input mismatch problem in this setting, by simulating inference-time generator samples during training time. Taken together, our improvements set new benchmarks in one-step image generation, with FID scores of 1.28 on ImageNet-64x64 and 8.35 on zero-shot COCO 2014, surpassing the original teacher despite a 500X reduction in inference cost. Further, we show our approach can generate megapixel images by distilling SDXL, demonstrating exceptional visual quality among few-step methods.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 19 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. One Step Diffusion via Shortcut Models

    cs.LG 2024-10 conditional novelty 7.0

    Shortcut models enable high-quality single or few-step sampling in diffusion models with one network and training phase by conditioning on desired step size.

  2. Dynamic-in-Few-Step: Unifying Dynamic Computation and Few-Step Distillation for Efficient Video Generation

    cs.CV 2026-07 conditional novelty 6.5

    Joint few-step distillation and step-specific structural pruning turns a video diffusion model into a compact Mixture-of-Models that cuts 24% extra FLOPs per step and reaches 30× speedup on Wan-14B.

  3. Wan-Streamer v0.1: End-to-end Real-time Interactive Foundation Models

    cs.CV 2026-06 unverdicted novelty 6.0

    Wan-Streamer is a unified end-to-end Transformer for low-latency streaming audio-visual interaction using block-causal attention on interleaved multimodal tokens.

  4. Flash-WAM: Modality-Aware Distillation for World Action Models

    cs.LG 2026-06 unverdicted novelty 6.0

    Flash-WAM introduces modality-specific consistency parametrizations to distill joint video-action diffusion models to single-step inference, delivering 23x speedup with preserved benchmark performance.

  5. RT-Lynx: Putting the GEMM Sparsity In a Right Way for Diffusion Models

    cs.LG 2026-05 unverdicted novelty 6.0

    RT-Lynx shifts DiT sparsity from weights to activations, reports up to 1.55x linear-layer speedup while preserving generation quality across multiple diffusion models.

  6. Variance Reduction for Expectations with Diffusion Teachers

    cs.LG 2026-05 unverdicted novelty 6.0

    CARV amortizes upstream diffusion teacher costs over noise resamples with timestep importance sampling and stratified-inverse-CDF sampling, delivering 2-3x effective compute gains in text-to-3D experiments and order-o...

  7. Unlocking Complex Visual Generation via Closed-Loop Verified Reasoning

    cs.CV 2026-05 unverdicted novelty 6.0

    CLVR couples verified logical planning with pixel diffusion, uses proxy reinforcement learning on distilled histories, and merges weights to cut inference to 4 NFEs while outperforming open-source T2I models on comple...

  8. Unlocking Complex Visual Generation via Closed-Loop Verified Reasoning

    cs.CV 2026-05 unverdicted novelty 6.0

    CLVR framework adds closed-loop visual verification, proxy prompt reinforcement learning, and delta-space weight merge to improve complex text-to-image generation over single-step or unverified multi-step baselines.

  9. LIVEditor-14B: Lightning Unified Video Editing via In-Context Sparse Attention

    cs.CV 2026-05 unverdicted novelty 6.0

    LIVEditor-14B applies a new sparse attention method (ISA) that prunes context and uses query-sharpness routing to cut attention latency ~60% with no loss in editing quality on standard benchmarks.

  10. DOLLAR: Few-Step Video Generation via Distillation and Latent Reward Optimization

    cs.CV 2024-12 unverdicted novelty 6.0

    DOLLAR combines variational score and consistency distillation for few-step video generation plus latent reward optimization, reporting 82.57 VBench score and up to 278x speedup over the teacher diffusion model for 12...

  11. Directing the World: Fast Autoregressive Video Generation with Compositional Human-Camera Control

    cs.CV 2026-06 unverdicted novelty 5.0

    A decoupled-control autoregressive video model using Fast-Slow Memory training, dynamic projection, and staged camera control to produce stable long-horizon outputs with human and viewpoint guidance.

  12. Wan-Streamer v0.1: End-to-end Real-time Interactive Foundation Models

    cs.CV 2026-06 unverdicted novelty 5.0

    Wan-Streamer is a unified Transformer model for low-latency streaming audio-visual interaction that jointly handles perception, reasoning, generation, and timing without external modules.

  13. Wan-Streamer v0.1: End-to-end Real-time Interactive Foundation Models

    cs.CV 2026-06 unverdicted novelty 5.0

    Wan-Streamer presents a unified end-to-end Transformer for low-latency multimodal streaming interaction without external modules.

  14. Ultra Flash: Scaling Real-Time Streaming Video Generation to High Resolutions

    cs.CV 2026-06 unverdicted novelty 5.0

    Ultra Flash introduces a cascaded streaming super-resolution framework with specialized training, upsampling, and optimization to enable real-time high-resolution video generation from low-res diffusion models.

  15. Variance Reduction for Expectations with Diffusion Teachers

    cs.LG 2026-05 unverdicted novelty 5.0

    CARV introduces a hierarchical Monte Carlo estimator with amortized reuse, importance sampling, and stratification that yields 2-3x effective compute gains on diffusion-teacher pipelines while cutting gradient varianc...

  16. Consistency Regularised Gradient Flows for Inverse Problems

    stat.ML 2026-05 unverdicted novelty 5.0

    A consistency-regularized Euclidean-Wasserstein-2 gradient flow performs joint posterior sampling and prompt optimization in latent space for efficient low-NFE inverse problem solving with diffusion models.

  17. LIVEditor-14B: Lightning Unified Video Editing via In-Context Sparse Attention

    cs.CV 2026-05 unverdicted novelty 5.0

    ISA prunes low-saliency context tokens and routes queries by sharpness to either full or 0-th order Taylor sparse attention, enabling LIVEditor to cut attention latency ~60% while beating prior video editing methods o...

  18. The Score-Difference Flow for Implicit Generative Modeling

    cs.LG 2023-04 unverdicted novelty 5.0

    Score-difference flow reduces KL divergence between distributions and is formally equivalent to denoising diffusion models and a hidden subproblem in optimal GAN training under stated conditions.

  19. ERNIE-Image Technical Report

    cs.CV 2026-05 unverdicted novelty 4.0

    The paper presents ERNIE-Image, an open-source 8B DiT text-to-image model claiming leading open-source performance and near-commercial results via specialized data construction and DPO alignment.