提出双输出生成框架，2步即达ImageNet最先进效果
Duality Models: An Embarrassingly Simple One-step Generation Paradigm
- 单输入双输出设计，同时预测速度与流映射
- 仅2步生成即达FID 1.79，显著提升效率
- 适合追求高效高质图像生成的研究者
基于一致性思想的生成模型如Shortcut和MeanFlow通过目标感知设计，在求解概率流ODE（PF-ODE）方面表现优异。传统方法引入目标时间 $r$ 与当前时间 $t$ 调控输出，从局部多步导数（$r = t$）到全局少步积分（$r = 0$）过渡。然而，常规的“一输入一输出”范式将训练预算分隔，通常将大量样本（如MeanFlow中的75%）分配给多步目标以保证稳定，导致少步生成训练不足，影响收敛并限制可扩展性。为此，我们提出杜阿尔模型（DuMo），采用“一输入双输出”范式：共享主干网络，双分支头同时从 $x_t$ 预测速度 $v_t$ 和流映射 $u_t$。该设计使多步目标的几何约束作用于每个样本，无需分离训练目标即可有效约束少步估计，大幅提升稳定性和效率。在256×256的ImageNet上，使用679M Diffusion Transformer与SD-VAE，仅需2步生成即达到最先进的FID 1.79。代码已开源。
原文摘要 · Abstract (English)
Consistency-based generative models like Shortcut and MeanFlow achieve impressive results via a target-aware design for solving the Probability Flow ODE (PF-ODE). Typically, such methods introduce a target time $r$ alongside the current time $t$ to modulate outputs between a local multi-step derivative ($r = t$) and a global few-step integral ($r = 0$). However, the conventional "one input, one output" paradigm enforces a partition of the training budget, often allocating a significant portion (e.g., 75% in MeanFlow) solely to the multi-step objective for stability. This separation forces a trade-off: allocating sufficient samples to the multi-step objective leaves the few-step generation undertrained, which harms convergence and limits scalability. To this end, we propose Duality Models (DuMo) via a "one input, dual output" paradigm. Using a shared backbone with dual heads, DuMo simultaneously predicts velocity $v_t$ and flow-map $u_t$ from a single input $x_t$. This applies geometric constraints from the multi-step objective to every sample, bounding the few-step estimation without separating training objectives, thereby significantly improving stability and efficiency. On ImageNet 256 $\times$ 256, a 679M Diffusion Transformer with SD-VAE achieves a state-of-the-art (SOTA) FID of 1.79 in just 2 steps. Code is available at: https://github.com/LINs-lab/DuMo
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。