New generative models leverage Wasserstein flows for faster, higher-quality outputs

报道来源 [7]

1. arXiv stat.ML TIER_1 English(EN) · Yu-Jui Huang, Zachariah Malik · 2026-05-14 04:00

   Generative Modeling by Minimizing the Wasserstein-2 Loss

   arXiv:2406.13619v4 Announce Type: replace Abstract: This paper develops a generative model by minimizing the second-order Wasserstein loss (the $W_2$ loss) through a distribution-dependent ordinary differential equation (ODE), whose dynamics involves the Kantorovich potential ass…

2. arXiv stat.ML TIER_1 English(EN) · Samuel Willis, Paul Duckworth, Jack Simons, Aleksandra Kalisz, Krisztina Sinkovics, Noam Ghenassia, Shikha Surana, Henry T. Oldroyd, Alexandru I. Stere, Dragos D Margineantu, Carl Henrik Ek, Henry Moss, Erik Bodin · 2026-05-14 04:00

   Sample-Efficient Optimisation over the Outputs of Generative Models

   arXiv:2509.23800v3 Announce Type: replace Abstract: Modern generative AI models, such as diffusion and flow matching models, can sample from rich data distributions. However, many applications, especially in science and engineering, require more than drawing samples from the mode…

3. arXiv stat.ML TIER_1 English(EN) · Romann M. Weber · 2026-05-14 04:00

   The Score-Difference Flow for Implicit Generative Modeling

   arXiv:2304.12906v4 Announce Type: replace-cross Abstract: Implicit generative modeling (IGM) aims to produce samples of synthetic data matching the characteristics of a target data distribution. Recent work (e.g. score-matching networks, diffusion models) has approached the IGM p…

4. arXiv stat.ML TIER_1 English(EN) · Daniel Paulin · 2026-05-14 02:20

   Training-Free Generative Sampling via Moment-Matched Score Smoothing

   Diffusion models generate samples by denoising along the score of a perturbed target distribution. In practice, one trains a neural diffusion model, which is computationally expensive. Recent work suggests that score matching implicitly smooths the empirical score, and that this …

5. arXiv stat.ML TIER_1 Deutsch(DE) · Jiaqi Han, Puheng Li, Qiushan Guo, Renyuan Xu, Stefano Ermon, Emmanuel J. Cand\`es · 2026-05-13 04:00

   One-Step Generative Modeling via Wasserstein Gradient Flows

   arXiv:2605.11755v1 Announce Type: cross Abstract: Diffusion models and flow-based methods have shown impressive generative capability, especially for images, but their sampling is expensive because it requires many iterative updates. We introduce W-Flow, a framework for training …

6. arXiv stat.ML TIER_1 Deutsch(DE) · Emmanuel J. Candès · 2026-05-12 08:29

   One-Step Generative Modeling via Wasserstein Gradient Flows

   Diffusion models and flow-based methods have shown impressive generative capability, especially for images, but their sampling is expensive because it requires many iterative updates. We introduce W-Flow, a framework for training a generator that transforms samples from a simple …

7. arXiv stat.ML TIER_1 English(EN) · Panos Tsimpos, Daniel Sharp, Youssef Marzouk · 2026-05-11 04:00

   One-Shot Generative Flows: Existence and Obstructions

   arXiv:2604.15439v3 Announce Type: replace Abstract: We study dynamic measure transport for generative modeling, focusing on transport maps that connect a source measure $P_0$ to a target measure $P_1$ by integrating a velocity field of the form $v_t(x) = \mathbb{E}[\dot X_t \mid …