arXiv subjects

Chuyu Liu

Publications and source records attributed to Chuyu Liu.

3 recordsLinked to original sources

arXiv2024

RATIONALYST: Mining Implicit Rationales for Process Supervision of Reasoning

The reasoning steps generated by LLMs might be incomplete, as they mimic logical leaps common in everyday communication found in their pre-training data: underlying rationales are frequently left implicit (unstated). To address this challenge, we introduce RATIONALYST, a model for process-supervision of reasoning based on pre-training on a vast collection of rationale annotations extracted from unlabeled data. We extract 79k rationales from web-scale unlabelled dataset (the Pile) and a combination of reasoning datasets with minimal human intervention. This web-scale pre-training for reasoning allows RATIONALYST to consistently generalize across diverse reasoning tasks, including mathematical, commonsense, scientific, and logical reasoning. Fine-tuned from LLaMa-3-8B, RATIONALYST improves the accuracy of reasoning by an average of 3.9% on 7 representative reasoning benchmarks. It also demonstrates superior performance compared to significantly larger verifiers like GPT-4 and similarly sized models fine-tuned on matching training sets.

cs.AI↗

arXiv2024

Graph Image Prior for Unsupervised Dynamic Cardiac Cine MRI Reconstruction

The inductive bias of the convolutional neural network (CNN) can be a strong prior for image restoration, which is known as the Deep Image Prior (DIP). Recently, DIP is utilized in unsupervised dynamic MRI reconstruction, which adopts a generative model from the latent space to the image space. However, existing methods usually use a pyramid-shaped CNN generator shared by all frames, embedding the temporal modeling within the latent space, which may hamper the model expression capability. In this work, we propose a novel scheme for dynamic MRI representation, named ``Graph Image Prior'' (GIP). GIP adopts a two-stage generative network in a new modeling methodology, which first employs independent CNNs to recover the image structure for each frame, and then exploits the spatio-temporal correlations within the feature space parameterized by a graph model. A graph convolutional network is utilized for feature fusion and dynamic image generation. In addition, we devise an ADMM algorithm to alternately optimize the images and the network parameters to improve the reconstruction performance. Experiments were conducted on cardiac cine MRI reconstruction, which demonstrate that GIP outperforms compressed sensing methods and other DIP-based unsupervised methods, significantly reducing the performance gap with state-of-the-art supervised algorithms. Moreover, GIP displays superior generalization ability when transferred to a different reconstruction setting, without the need for any additional data.

eess.IV↗

arXiv2023

Accelerated partial separable model using dimension-reduced optimization technique for ultra-fast cardiac MRI

Objective. Imaging dynamic object with high temporal resolution is challenging in magnetic resonance imaging (MRI). Partial separable (PS) model was proposed to improve the imaging quality by reducing the degrees of freedom of the inverse problem. However, PS model still suffers from long acquisition time and even longer reconstruction time. The main objective of this study is to accelerate the PS model, shorten the time required for acquisition and reconstruction, and maintain good image quality simultaneously. Approach. We proposed to fully exploit the dimension reduction property of the PS model, which means implementing the optimization algorithm in subspace. We optimized the data consistency term, and used a Tikhonov regularization term based on the Frobenius norm of temporal difference. The proposed dimension-reduced optimization technique was validated in free-running cardiac MRI. We have performed both retrospective experiments on public dataset and prospective experiments on in-vivo data. The proposed method was compared with four competing algorithms based on PS model, and two non-PS model methods. Main results. The proposed method has robust performance against shortened acquisition time or suboptimal hyper-parameter settings, and achieves superior image quality over all other competing algorithms. The proposed method is 20-fold faster than the widely accepted PS+Sparse method, enabling image reconstruction to be finished in just a few seconds. Significance. Accelerated PS model has the potential to save much time for clinical dynamic MRI examination, and is promising for real-time MRI applications.

eess.IV↗