Next-Acceleration-Scale Prediction for Autoregressive MRI Reconstruction

Hugging Face Daily Papers Papers

Summary

Discrete autoregressive MRI reconstruction using privileged information distillation achieves superior performance under extreme undersampling by leveraging visual autoregressive modeling techniques.

MRI reconstruction is an inherently ill-posed inverse problem, since incomplete measurements admit many plausible solutions. This ambiguity becomes more severe under high acceleration, where pixel-domain continuous predictors tend to average over feasible reconstructions and suppress high-frequency anatomy. We address this limitation by moving reconstruction to discrete multi-scale latent space and posing it as autoregressive next-acceleration-scale prediction. Leveraging discrete priors proven effective in visual autoregressive modeling, our method restricts the solution to compact sequences of codebook tokens, enabling sharp reconstructions even from extremely sparse measurements. This discrete autoregressive formulation also aligns naturally with modern large language model post-training techniques. Building on this observation, we introduce on-policy privileged information distillation for visual autoregressive modeling, where a teacher is provided training only privileged context that is unavailable at inference, in our case fully sampled acquisitions, and supervises a student trained on its own rollouts, leading to consistent reconstruction gains. Through extensive experiments on the fastMRI benchmark, we show that our approach delivers improved reconstruction performance across diverse sampling patterns under extreme undersampling. Project website is https://yilmazkorkmaz1.github.io/discrete-mri-reconstruction-opd/{here}.
Original Article
View Cached Full Text

Cached at: 05/25/26, 10:40 PM

Paper page - Next-Acceleration-Scale Prediction for Autoregressive MRI Reconstruction

Source: https://huggingface.co/papers/2605.19354

Abstract

Discrete autoregressive MRI reconstruction using privileged information distillation achieves superior performance under extreme undersampling by leveraging visual autoregressive modeling techniques.

MRI reconstruction is an inherently ill-posed inverse problem, since incomplete measurements admit many plausible solutions. This ambiguity becomes more severe under high acceleration, where pixel-domain continuous predictors tend to average over feasible reconstructions and suppress high-frequency anatomy. We address this limitation by moving reconstruction todiscrete multi-scale latent spaceand posing it asautoregressive next-acceleration-scale prediction. Leveraging discrete priors proven effective invisual autoregressive modeling, our method restricts the solution to compact sequences ofcodebook tokens, enabling sharp reconstructions even from extremely sparse measurements. This discrete autoregressive formulation also aligns naturally with modern large language model post-training techniques. Building on this observation, we introduceon-policyprivileged information distillationforvisual autoregressive modeling, where a teacher is provided training only privileged context that is unavailable at inference, in our case fully sampled acquisitions, and supervises a student trained on its own rollouts, leading to consistent reconstruction gains. Through extensive experiments on thefastMRI benchmark, we show that our approach delivers improved reconstruction performance across diverse sampling patterns underextreme undersampling. Project website is https://yilmazkorkmaz1.github.io/discrete-mri-reconstruction-opd/{here}.

View arXiv pageView PDFProject pageGitHub1Add to collection

Get this paper in your agent:

hf papers read 2605\.19354

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.19354 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.19354 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.19354 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

PhyMRI-SR: Toward Physics-Aware MRI Image Super-Resolution

Hugging Face Daily Papers

This paper proposes PhyMRI-SR, a physics-aware MRI super-resolution method that uses Gaussian splatting and physics-constrained modeling to dynamically adapt resolution-SNR configurations, achieving state-of-the-art performance.