New research advances Rectified Flow and Flow Matching generative models · 10 sources tracked
ByPulseAugur Editorial·[15 sources]·
Researchers are exploring new methods within the framework of Rectified Flow (RF) and Flow Matching (FM) generative models. One approach, Self-Consistent Flow (SC-Flow), unifies velocity and endpoint prediction to stabilize training and improve generation quality. Another area of investigation focuses on understanding and mitigating memorization in RF models, with new metrics developed to detect privacy risks. Additionally, techniques like Velocity Scheduled Flow Matching (VSFM) are being introduced to enhance sampling efficiency and generation fidelity by adjusting the speed profile during inference and training. Finally, Reward Transport offers a novel way to control generated properties by aligning noise and data based on target attributes.
AI
IMPACT
These advancements in Flow Matching and Rectified Flow models could lead to more efficient, higher-fidelity generative models with better control over data properties and reduced memorization risks.
RANK_REASON
Multiple arXiv papers introducing novel methods and analyses within the Flow Matching and Rectified Flow generative modeling frameworks.
arXiv:2607.12171v1 Announce Type: cross Abstract: In rectified-flow-based generative models, the neural network can be trained to predict two different targets, such as the instantaneous velocity or the data endpoint, to perform denoising. Although prior work shows that these par…
arXiv cs.LG
TIER_1English(EN)·Mingxing Rao, Daniel Moyer·
arXiv:2603.13421v2 Announce Type: replace Abstract: Generative models based on the Flow Matching objective, particularly Rectified Flow, have emerged as a dominant paradigm for efficient, high-fidelity image synthesis. However, while existing research heavily prioritizes generati…
arXiv cs.LG
TIER_1English(EN)·Gianluca Galletti, Gerald Gutenbrunner, William Hornsby, Lorenzo Zanisi, Naomi Carey, Stanislas Pamela, Johannes Brandstetter, Fabian Paischer·
arXiv:2607.13022v1 Announce Type: cross Abstract: Many nonlinear physical systems exhibit an initial transient phase in which perturbations grow before nonlinear interactions lead to a statistically steady state. While this saturated regime is of primary interest, direct numerica…
arXiv:2607.12616v1 Announce Type: new Abstract: Continuous-time generative frameworks construct probability paths between base and target domains by optimizing time-dependent velocity fields. While theoretical targets favor straight trajectories, empirical networks develop comple…
Many nonlinear physical systems exhibit an initial transient phase in which perturbations grow before nonlinear interactions lead to a statistically steady state. While this saturated regime is of primary interest, direct numerical simulations must resolve the full transient dyna…
Many nonlinear physical systems exhibit an initial transient phase in which perturbations grow before nonlinear interactions lead to a statistically steady state. While this saturated regime is of primary interest, direct numerical simulations must resolve the full transient dyna…
Continuous-time generative frameworks construct probability paths between base and target domains by optimizing time-dependent velocity fields. While theoretical targets favor straight trajectories, empirical networks develop complex path deformations. This paper presents the Fin…
Continuous-time generative frameworks construct probability paths between base and target domains by optimizing time-dependent velocity fields. While theoretical targets favor straight trajectories, empirical networks develop complex path deformations. This paper presents the Fin…
arXiv:2607.11442v1 Announce Type: new Abstract: Flow matching trains a neural network to regress the conditional velocity along a linear interpolant between noise and data, and the number of network evaluations~(NFE) sets the cost of sampling. The straight-line interpolant carrie…
Flow matching trains a neural network to regress the conditional velocity along a linear interpolant between noise and data, and the number of network evaluations~(NFE) sets the cost of sampling. The straight-line interpolant carries an implicit choice: the sample moves at consta…
Flow matching trains a neural network to regress the conditional velocity along a linear interpolant between noise and data, and the number of network evaluations~(NFE) sets the cost of sampling. The straight-line interpolant carries an implicit choice: the sample moves at consta…
arXiv cs.AI
TIER_1English(EN)·Kehan Guo, Yili Shen, Yujun Zhou, Yue Huang, Chujie Gao, Shiyi Du, Xiangliang Zhang·
arXiv:2607.08781v1 Announce Type: cross Abstract: The coupling in flow matching -- the rule pairing noise vectors with data points -- is typically treated as a computational choice. We show that this coupling can instead serve as an alignment interface: by matching noise and data…
Flow matching is a powerful tool for generative modeling, but emerging applications in robotics, planning, and physics require inference-time constraints on generated outputs. Such constraints are often complex and highly nonlinear. As a result, methods designed for linear constr…
arXiv:2607.13841v1 Announce Type: cross Abstract: Heavy-tailed data arise in many domains where rare events carry disproportionate importance, such as imbalanced image datasets, financial returns, and weather extremes. Standard diffusion and flow-matching models typically begin f…
Heavy-tailed data arise in many domains where rare events carry disproportionate importance, such as imbalanced image datasets, financial returns, and weather extremes. Standard diffusion and flow-matching models typically begin from Gaussian noise or Gaussian source distribution…
