# TFTF: Training-Free Targeted Flow for Conditional Sampling Source: [TFTF: Training-Free Targeted Flow for Conditional Sampling]() Collection index: [llms.txt]() Data updated: 2026-02-17T05:00:04.915Z Authors: Qianqian Qu, Jun S. Liu Published: 2026-02-13T13:41:35.000Z ## Abstract We propose a training-free conditional sampling method for flow matching models based on importance sampling. Because a na\\"ive application of importance sampling suffers from weight degeneracy in high-dimensional settings, we modify and incorporate a resampling technique in sequential Monte Carlo (SMC) during intermediate stages of the generation process. To encourage generated samples to diverge along distinct trajectories, we derive a stochastic flow with adjustable noise strength to replace the deterministic flow at the intermediate stage. Our framework requires no additional training, while providing theoretical guarantees of asymptotic accuracy. Experimentally, our method significantly outperforms existing approaches on conditional sampling tasks for MNIST and CIFAR-10. We further demonstrate the applicability of our approach in higher-dimensional, multimodal settings through text-to-image generation experiments on CelebA-HQ. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Class-conditional sampling on MNIST (test)](): Class-conditional sampling; MNIST (test); Accuracy (ACCL): 99.66. - [Class-conditional sampling on CIFAR-10](): Class-conditional sampling; CIFAR-10; Conditional Accuracy (Label): 92.56. - [Text-to-image Generation on CelebA-HQ](): Text-to-Image Generation; CelebA-HQ; CLIP\_L: 0.295. ## Original sources and implementation [Original paper]() [PDF]()