arXiv：cs.AI· Sylvain Dehayem Kenfouo, Aaron Zweig, Mingxuan Zhang, Elham Azizi·· 16 小时前
无需训练的可控多样采样：利用回拉方向提升扩散模型生成多样性
Controllable Diverse Sampling through Pullback Directions
AI 导读
研究者提出一种无需训练的方法，利用条件嵌入空间到图像空间的映射所诱导的局部回拉几何，识别输出最敏感的方向，沿这些方向引导不同采样轨迹，从而在保持与条件一致的前提下生成更多样的结果。在 MS-COCO 和 BrushNet 上的实验表明，该方法能提升条件图像生成与图像修复的多样性，且无需额外优化或采样开销。该工作已被 NeurIPS 2026 的 DynaFront workshop 接收。
正文
Abstract:Conditional diffusion models can generate high-quality images that align closely with a given prompt. However, samples generated from different random seeds may remain visually similar, limiting creative exploration. Existing approaches can increase variation, but often require additional optimization or sampling overhead, while stronger perturbations may introduce distortions that degrade image quality or prompt alignment. We propose a training-free method that leverages the local pullback geometry induced by the mapping from the conditioning-embedding space to the image space to identify directions of maximal output sensitivity. By steering different sampling trajectories along these directions, we encourage distinct generations while maintaining consistency with the conditioning signal. Experiments on MS-COCO and BrushNet show the efficacy of the method in improving diversity on prompt-conditioned image generation and inpainting.
| Comments: | Accepted at the DynaFront workshop at NeurIPS 2026 |
| Subjects: | Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2610.14294 [cs.AI] |
| (or arXiv:2610.14294v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.14294 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sylvain Dehayem Kenfouo [view email]
[v1]
Fri, 9 Oct 2026 20:17:27 UTC (12,585 KB)
来源：arXiv：cs.AI · arxiv.org