提出一种新生成模型，8步即可生成高质量图像，无需复杂训练。
Inductive Moment Matching
- 通过单阶段训练直接学习采样分布，不依赖预训练模型。
- 在ImageNet上仅用8步推理就达到1.99的FID，CIFAR-10两步达1.98。
- 对超参数和模型结构鲁棒，适合快速部署的生成任务。
扩散模型与流匹配生成高质量样本，但推理速度慢，将其蒸馏为少步模型常导致不稳定且需大量调参。为解决这一权衡，我们提出归纳矩匹配（IMM），一种支持单步或少步采样的新型生成模型，采用单阶段训练。与蒸馏不同，IMM无需预训练初始化和双网络优化；与一致性模型相比，IMM保证分布收敛，在多种超参数和标准架构下保持稳定。IMM在ImageNet-256x256上仅用8步推理即达到1.99 FID，于从零训练的模型在CIFAR-10上实现2步1.98的领先FID。
原文摘要 · Abstract (English)
Diffusion models and Flow Matching generate high-quality samples but are slow at inference, and distilling them into few-step models often leads to instability and extensive tuning. To resolve these trade-offs, we propose Inductive Moment Matching (IMM), a new class of generative models for one- or few-step sampling with a single-stage training procedure. Unlike distillation, IMM does not require pre-training initialization and optimization of two networks; and unlike Consistency Models, IMM guarantees distribution-level convergence and remains stable under various hyperparameters and standard model architectures. IMM surpasses diffusion models on ImageNet-256x256 with 1.99 FID using only 8 inference steps and achieves state-of-the-art 2-step FID of 1.98 on CIFAR-10 for a model trained from scratch.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。