自洽流：统一整流流模型的速度和端点预测
Self-Consistent Flow: Unifying Velocity and Endpoint Prediction for Rectified Flow Models
浏览论文内容
中文总结 AI 辅助
研究基于整流流的生成模型中不同参数化预测目标的问题，提出自洽流方法，通过轻量级一致性损失联合训练网络预测速度和端点，提升模型性能，在图像生成任务中显著优于标准整流流基线。
中文摘要 AI 辅助
在基于整流流的生成模型中，神经网络可训练预测瞬时速度或数据端点等不同目标来执行去噪。尽管先前工作表明这些参数化导致不同经验行为，但其各自优势的潜在机制仍未充分探索，如何有效结合也不清楚。本文分析不同参数化的学习误差如何影响生成性能，发现预测数据端点有稳定训练的清晰信号，预测速度能维持数据流形附近稳定采样动态。基于此提出自洽流（SC-Flow），通过轻量级一致性损失联合训练单个网络预测局部速度和数据端点，两者预测的一致性提升了模型性能。该方法无需重大架构改变且计算开销极小。在图像生成任务上的大量实验表明，SC-Flow显著稳定了优化并改善了生成路径的直线度，相比标准整流流基线在生成质量上有显著提升。
英文摘要
In rectified-flow-based generative models, the neural network can be trained to predict two different targets, such as the instantaneous velocity or the data endpoint, to perform denoising. Although prior work shows that these parameterizations lead to different empirical behaviors, the mechanisms underlying their respective advantages remain to be underexplored, and how to combine them effectively is still unclear. In this work, we analyze how learning errors from different parameterizations affect the generation performance. We show that predicting the data endpoint has a clear training signal that stabilizes training, whereas predicting the velocity maintains stable sampling dynamics near the data manifold. Motivated by these insights, we propose Self-Consistent Flow (SC-Flow), a new method that unifies the benefits of both parameterizations. By employing a lightweight consistency loss, SC-Flow jointly trains a single network to predict both the local velocity and the data endpoint, and the consistency between the two predictions improves the model's performance. The method requires no major architectural changes and adds minimal computational overhead. Extensive experiments on image generation tasks demonstrate that SC-Flow substantially stabilizes optimization and improves the straightness of generation paths, leading to significant gains in generation quality over standard rectified-flow baselines.
发表机构
- Tufts University（塔夫茨大学）
机构由 AI 辅助整理，请以论文原文为准。
