arXiv：cs.LG· Eliot Beyler, Francis Bach·· 6 小时前
Weighted Adjoint Matching：用时间相关控制成本微调流匹配模型
Weighted Adjoint Matching: Finetuning Flow Matching Models with Time-Dependent Control Cost
AI 导读
Weighted Adjoint Matching 通过在控制成本上引入时间相关权重，对预训练的扩散或流匹配模型进行微调。该方法在生成早期施加高成本以保持粗粒度特征的平衡与多样性，后期低成本则让模型细化细节以提升奖励。在 Gaussian mixtures 和 MNIST 上，其奖励与模式平衡的权衡优于 Adjoint Matching。
正文
Abstract:We introduce Weighted Adjoint Matching, a method to finetune a pretrained diffusion or flow matching model given a differentiable reward. We derive it from Adjoint Matching, which reformulates finetuning as a stochastic optimal control problem with a quadratic control cost. Building on the observation that these models generate coarse features before fine details, we place a time-dependent weight on the control cost. A high cost early in the generative process preserves balance and diversity in the coarse features, which carry most of the semantics, while a low cost at the end lets the model increase the reward by refining details. We prove that our method is equivalent to maximizing the reward under time-varying Kullback--Leibler regularization on the generative paths. Experiments on Gaussian mixtures and on MNIST show a better tradeoff between reward and mode balance than Adjoint Matching.
| Subjects: | Machine Learning (stat.ML); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.20035 [stat.ML] |
| (or arXiv:2610.20035v1 [stat.ML] for this version) | |
| https://doi.org/10.48550/arXiv.2610.20035 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Eliot Beyler [view email]
[v1]
Wed, 14 Oct 2026 11:43:27 UTC (7,340 KB)
来源：arXiv：cs.LG · arxiv.org