arXiv cs.AI· Jinjian Jiang, Jing Peng, Rehan Deen, Yu Cao, Ziquan Liu, Muhammad Salman Haleem, David Mguni·· 11 小时前AI 评分27
Meta-Diff：用源信息任务空间迁移实现少样本扩散模型适配
Bridging the Data Gap: Source-Informed Task-Space Transport for Few-Shot Diffusion
AI 导读
Meta-Diff 将源域与目标域分布嵌入共享低维任务空间，学习配对训练任务间的仿射迁移映射，并借助共享扩散主干从迁移后的源坐标生成，部署时只需源样本和已学关系标识，无需目标图像或目标端优化。
正文 · 原文
Abstract:When an object category is well represented in one domain but scarcely observed in another, adapting a diffusion model to the target can be unreliable. Yet the same source-target relationship may have been observed for other categories. We ask whether this recurring relation can be learned and reused for a new semantic task. We introduce Meta-Diff, which embeds source and target distributions in a shared low-dimensional task space, learns an affine transport map across paired training tasks, and generates from the transported source coordinate using a shared diffusion backbone. At deployment, it requires source examples and the identity of a previously learned relation, but neither target images nor target-specific optimisation. Gaussian-mixture experiments examine how transfer depends on task information retained by the encoder, relation predictability and the size of the source--target shift. On held-out CIFAR-100 validation tasks, correct-source transport achieves 13.9% lower mean sliced Wasserstein distance than wrong-source transport. In the one-target final-test comparison, it achieves 14.7% lower mean distance than target-only adaptation and 32.7% lower than full fine-tuning under validation-selected settings. These gains have clear limits: directly applying the known corruption to clean-source generations performs better on CIFAR, and requested-class identity remains weak. Office-Home shows domain alignment without reliable object identity; clinical stress tests show no resolved source-specific advantage. The results characterise when reusable task-space relations help and where their generative limitations remain.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.31323 [cs.LG] |
| (or arXiv:2610.31323v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.31323 arXiv-issued DOI via DataCite |
Submission history
From: David Mguni [view email]
[v1]
Thu, 22 Oct 2026 14:13:59 UTC (7,532 KB)
[v2]
Wed, 4 Nov 2026 14:42:39 UTC (7,188 KB)
来源：arXiv cs.AI · arxiv.org