aiqbench analysis
MPC-Flow addresses the computational bottleneck of guiding flow-based generative models by reframing trajectory optimization as a sequence of control sub-problems, enabling practical inference-time guidance without backpropagation through the model dynamics. The approach demonstrates strong scalability on massive architectures like FLUX.2 (32B) using training-free methods on consumer hardware, offering a viable path for deploying high-fidelity image restoration tasks such as in-painting and super-resolution.
Solving Inverse Problems with Flow-based Models via Model Predictive Control
arXiv:2601.23231v2 Announce Type: replace-cross Abstract: Flow-based generative models provide strong unconditional priors for inverse problems, but guiding their dynamics for conditional generation remains challenging. Recent work casts training-free conditional generation in flow models as an optimal control problem; however, solving the resulting trajectory optimisation is computationally and memory intensive, requiring differentiation through the flow dynamics or adjoint solves. We propose MPC-Flow, a model predictive control framework that formulates inverse problem solving with flow-based generative models as a sequence of control sub-problems, enabling practical optimal control-based guidance at inference time. We provide theoretical analysis linking MPC-Flow to the underlying optimal control objective and show how different algorithmic choices yield a spectrum of guidance algorithms, including regimes that avoid backpropagation through the generative model trajectory. We evaluate MPC-Flow on benchmark image restoration tasks, spanning linear and non-linear settings such as in-painting, deblurring, and super-resolution, and demonstrate strong performance and scalability to massive state-of-the-art architectures via training-free guidance of FLUX.2 (32B) in a quantised setting on consumer hardware.
Abstract:Flow-based generative models provide strong unconditional priors for inverse problems, but guiding their dynamics for conditional generation remains challenging. Recent work casts training-free conditional generation in flow models as an optimal control problem; however, solving the resulting trajectory optimisation is computationally and memory intensive, requiring differentiation through the flow dynamics or adjoint solves. We propose MPC-Flow, a model predictive control framework that formulates inverse problem solving with flow-based generative models as a sequence of control sub-problems, enabling practical optimal control-based guidance at inference time. We provide theoretical analysis linking MPC-Flow to the underlying optimal control objective and show how different algorithmic choices yield a spectrum of guidance algorithms, including regimes that avoid backpropagation through the generative model trajectory. We evaluate MPC-Flow on benchmark image restoration tasks, spanning linear and non-linear settings such as in-painting, deblurring, and super-resolution, and demonstrate strong performance and scalability to massive state-of-the-art architectures via training-free guidance of FLUX.2 (32B) in a quantised setting on consumer hardware.
Image and Video Processing (eess.IV); Machine Learning (cs.LG)
arXiv:2601.23231 [eess.IV]
(or arXiv:2601.23231v2 [eess.IV] for this version)
https://doi.org/10.48550/arXiv.2601.23231 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 257 words
Related coverage
Physics-Guided Flow Matching for CT Image Reconstruction
arXiv · Sep 1, 2026
Trajectory Stitching for Solving Inverse Problems with Flow-Based Models
arXiv · Oct 2, 2026
Exact Posterior Score Estimation for Solving Linear Inverse Problems
arXiv · Jun 16, 2026
MedForj: An open, large-scale foundational generative prior for high-resolution 3D brain MRI
arXiv · Oct 7, 2026