通过可控制的生成场实现高质量图像合成与语义编辑。
Generative Anchored Fields: Controlled Data Generation via Emergent Velocity Fields and Transport Algebra
- 分解噪声与数据预测，从线性桥接中学习速度场进行生成。
- 在ImageNet和CelebA-HQ上达到FID 7.51和7.27，5-8步完成高质量采样。
- 支持多类别组合与语义编辑，适合需要可控生成的研究者。
我们提出生成锚定场（Generative Anchored Fields, GAF），一种从线性桥接任意点学习独立终点预测器 $J$（噪声）和 $K$（数据）的生成模型。与仅使用单一轨迹或评分预测的方法不同，GAF通过坐标学习直接恢复桥接端点。速度场 $v=K-J$ 由其时序条件下的不一致性自然涌现。该分解实现「传输代数」：对多个 $J/K$ 头进行代数操作，实现组合控制。通过类特定 $K_n$ 头，GAF定义共享基础噪声分布与多个数据域之间的定向传输映射，支持可控插值、多类别组合与语义编辑。可通过迭代终点精炼（IER）直接在预测数据坐标（$K$）上实现，采样仅需5-8步即达高质量生成；也可作用于涌现速度场（$v$）。在ImageNet $256 imes256$ 和 CelebA-HQ $256 imes256$ 上分别取得FID 7.51和7.27，无需分类器自由引导。代码已公开于https://github.com/IDLabMedia/GAF。
原文摘要 · Abstract (English)
We present Generative Anchored Fields (GAF), a generative model that learns independent endpoint predictors, $J$ (noise) and $K$ (data), from any point on a linear bridge. Unlike existing approaches that use a single trajectory or score predictor, GAF is trained to recover the bridge endpoints directly via coordinate learning. The velocity field $v=K-J$ emerges from their time-conditioned disagreement. This factorization enables \textit{Transport Algebra}: algebraic operations on multiple $J/K$ heads for compositional control. With class-specific $K_n$ heads, GAF defines directed transport maps between a shared base noise distribution and multiple data domains, allowing controllable interpolation, multi-class composition, and semantic editing. This is achieved either directly on the predicted data coordinates ($K$) using Iterative Endpoint Refinement (IER), a novel sampler that achieves high-quality generation in $5-8$ steps, or on the emergent velocity field ($v$). We achieve strong sample quality (FID 7.51 on ImageNet $256\times256$ and $7.27$ on CelebA-HQ $256\times 256$, without classifier-free guidance) while treating compositional generation as an architectural primitive. Code available at https://github.com/IDLabMedia/GAF.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。