Apple 提出归一化轨迹模型 NTM：四步采样保留精确似然

AI 摘要

Apple 机器学习研究团队（联合宾夕法尼亚大学、UIUC）提出 Normalizing Trajectory Models（NTM），将扩散模型的每个反向步骤建模为具有精确似然训练的条件归一化流，结合浅层可逆块与跨轨迹的深度并行预测器，可从头训练或从预训练流匹配模型初始化。其精确轨迹似然支持自蒸馏，使轻量去噪器仅用四步即可生成高质量样本，在文生图基准上四步采样即可匹配或超越强基线，并独特地保留生成轨迹的精确似然。

正文节选

Diffusion-based models decompose sampling into many small Gaussian denoising steps, an assumption that breaks down when generation is compressed to a few coarse transitions. Existing few-step methods address this through distillation, consistency training, or adversarial objectives, but sacrifice the likelihood framework in the process. We introduce Normalizing Trajectory Models (NTM), which models each reverse step as an expressive conditional normalizing flow with exact likelihood training. Ar

发布时间：2026-10-08 08:00

抓取时间：2026-10-13 07:57

来源机构：Apple