AegisFlow: Training-Free Non-Myopic Path-Safe Guided Flow Matching
Pretrained generative models, such as diffusion and flow matching models, have demonstrated exceptional capabilities in synthesizing high-quality data. However, steering their generative trajectories…
- 用途
- 原典で確認
- 難易度の目安
- Hard
- 実装コストの目安
- High
概要
Pretrained generative models, such as diffusion and flow matching models, have demonstrated exceptional capabilities in synthesizing high-quality data. However, steering their generative trajectories to satisfy strict constraints--such as identity preservation in image editing or obstacle avoidance in robotics--remains fundamentally challenging. While recent training-free guidance methods enforce path-safety, they rely heavily on min-norm controllers that only account for immediate constraint satisfaction. These myopic approaches often yield suboptimal trajectories that hover dangerously close to the boundary, severely restricting the model's ability to optimize toward its target. In this paper, we propose AegisFlow, a training-free, non-myopic path-safe guidance framework that recasts guided generation as a safety-critical optimal control problem with global scope. By integrating a time-varying barrier function directly into the global cost functional, AegisFlow naturally enforces constraints along the entire generative path. To compute the global optimal control, we derive a tractable, mathematically grounded approximation of the Hamilton-Jacobi-Bellman (HJB) equation. Comprehensive experiments demonstrate that AegisFlow outperforms state-of-the-art methods across text-guided image manipulation and robotic planning, establishing a significantly better trade-off between target alignment and constraint satisfaction. Code is available at https://github.com/LiuKunpeng03/AegisFlow.
参考リンク
リンク先のライセンス・再現性・データ条件をご確認ください。自動分類や実装の目安は保証ではありません。
実装を始めるプロンプト
参考リンクを確認したうえで、次の内容をコピーして利用できます。
# 論文・記事 AegisFlow: Training-Free Non-Myopic Path-Safe Guided Flow Matching # 概要 Pretrained generative models, such as diffusion and flow matching models, have demonstrated exceptional capabilities in synthesizing high-quality data. However, steering their generative trajectories… # 原典 https://arxiv.org/abs/2610.39825 上記の手法を実装してください。実行前に依存関係とライセンスを確認してください。