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 2602.12233cs.LG
收藏

分类流映射

Categorical Flow Maps

UvA-Bosch Delta Lab, University of Amsterdam, Amsterdam, Netherlands(阿姆斯特丹大学);Department of Computer Science, University of Oxford, Oxford, UK(牛津大学计算机科学系);Donders Institute for Brain, Cognition and Behaviour, Radboud University(拉德堡德大学大脑与行为研究所)

Daan Roos, Oscar Davis, Floor Eijkelboom, Michael Bronstein, Max Welling, İsmail İlkan Ceylan, Luca Ambrogioni, Jan-Willem van de Meent

AI总结:

分类流映射通过自我蒸馏实现加速的少步生成,适用于图像、分子图和文本,取得最佳性能。

摘要:

我们介绍了分类流映射,一种用于通过自我蒸馏加速生成类别数据的流匹配方法。在最近的变分流匹配公式和更广泛的趋势向扩散和基于流的模型中加速推理的基础上,我们定义了一个流向单纯形的流映射,将概率质量向预测终点运输,从而得到一种自然约束模型预测的参数化方法。由于我们的轨迹是连续而非离散的,分类流映射可以使用现有的蒸馏技术进行训练,以及基于终点一致性的新目标。这种连续公式也自动解锁了测试时推理:我们可以在类别设置中直接重用现有的指导和重新加权技术,以引导采样向下游目标。实验上,我们在图像、分子图和文本上实现了最先进的少步结果,即使在单步生成中也表现出色。

英文摘要:

We introduce Categorical Flow Maps, a flow-matching method for accelerated few-step generation of categorical data via self-distillation. Building on recent variational formulations of flow matching and the broader trend towards accelerated inference in diffusion and flow-based models, we define a flow map towards the simplex that transports probability mass toward a predicted endpoint, yielding a parametrisation that naturally constrains model predictions. Since our trajectories are continuous rather than discrete, Categorical Flow Maps can be trained with existing distillation techniques, as well as a new objective based on endpoint consistency. This continuous formulation also automatically unlocks test-time inference: we can directly reuse existing guidance and reweighting techniques in the categorical setting to steer sampling toward downstream objectives. Empirically, we achieve state-of-the-art few-step results on images, molecular graphs, and text, with strong performance even in single-step generation.

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。

↑