arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.20155cs.LGcs.CVstat.ML
收藏

超越单个标记:通过离散MMD蒸馏离散扩散模型

Beyond Single Tokens: Distilling Discrete Diffusion Models via Discrete MMD

Google DeepMind Amsterdam(谷歌深思亚马逊)

Emiel Hoogeboom, David Ruhe, Jonathan Heek, Thomas Mensink, Tim Salimans

AI总结:

本文提出D-MMD方法,通过离散MMD蒸馏离散扩散模型,实现高质量和多样性的生成,优于传统方法。

摘要:

目前难以蒸馏离散扩散模型,而连续扩散文献中有许多蒸馏方法可将采样步骤减少到少数。我们的方法,离散矩匹配蒸馏(D-MMD),借鉴了在连续领域中高度成功的理念。与之前离散蒸馏方法不同,D-MMD在足够采样步骤下保持高质量和多样性。这在文本和图像数据集上得到验证。此外,新蒸馏的生成器可超越其教师模型。

英文摘要:

It is currently difficult to distill discrete diffusion models. In contrast, continuous diffusion literature has many distillation approaches methods that can reduce sampling steps to a handful. Our method, Discrete Moment Matching Distillation (D-MMD), leverages ideas that have been highly successful in the continuous domain. Whereas previous discrete distillation methods collapse, D-MMD maintains high quality and diversity (given sufficient sampling steps). This is demonstrated on both text and image datasets. Moreover, the newly distilled generators can outperform their teachers.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑