arXiv subjects

Bowei Liu

Publications and source records attributed to Bowei Liu.

16 recordsLinked to original sources

arXiv2026

LV-CARE-Diff: A Conditional Anatomy-Aware Diffusion Model for Left Ventricular Shape Reconstruction and Function Quantification from Ultra-Sparse Cine Slices

Left ventricular functional quantification is an essential examination and is routinely performed using cardiovascular magnetic resonance (CMR) cine imaging. However, conventional CMR cine protocols require the acquisition of multiple short-axis (SAX) slices to cover the entire left ventricle (LV) along with two long-axis (LAX) slices, which is time-consuming and places a considerable burden on patients who are unable to sustain repeated breath-holds, limiting its suitability for large-scale early screening. In this study, a Conditional Anatomy-Aware Diffusion Model (LV-CARE-Diff) was developed using a coarse-to-fine strategy to reconstruct the complete LV shape from ultra-sparse cine slices, namely three short-axis and two long-axis slices, with the aim of accelerating CMR cine examination. LV-CARE-Diff employs a 3DUNet to generate a coarse initial shape, which is subsequently refined through a residual diffusion model. A condition-guided input incorporating imaging plane orientation and positional metadata was constructed to enable spatial awareness, and a multi-objective training strategy jointly supervising shape, function, and anatomy was incorporated to guide high-fidelity reconstruction. LV-CARE-Diff was compared against a standalone 3DUNet, a standalone diffusion model, and a 3D UNet with diffusion-based refinement. Testing results indicated that complete LV shape could be robustly reconstructed by all deep learning models, with the highest reconstruction performance achieved by the proposed LV-CARE-Diff. Deep learning models reconstructing LV shape from sparse cine slices preserved 96% of functional quantification accuracy while reducing imaging time by 73%. The LV-CARE-Diff framework established in this study enables ultra-sparse cine acquisition to shorten CMR examination duration without sacrificing quantitative functional accuracy.

physics.med-ph↗

arXiv2026

VidForensics-M1: Meta-Detection Reinforcement Learning with Verifiable Temporal Grounding for AI-Generated Video Forensics

Recent advances in video generation models have significantly improved the realism of synthetic videos, blurring the boundary between generated and authentic content and raising concerns about misinformation. Existing MLLM-based detectors mainly rely on supervised fine-tuning or label-level reinforcement learning, where coarse supervision limits generalization to unseen scenarios and emerging video generators. To overcome these limitations, we are the first to introduce \textbf{meta-detection} into AI-generated video detection, enabling reliable forgery detection by jointly optimizing predicted labels and supporting evidence within reinforcement learning. This paradigm requires reliable evidence signals and effective mechanisms to integrate them into label-level optimization. Textual rationales provide semantic descriptions of forgery artifacts, but their generation and verification depend on external models, making supervision vulnerable to hallucinations and semantic biases. In contrast, temporal grounding provides more objective and verifiable evidence, as manipulated intervals can be precisely controlled during forgery construction. Based on this insight, we propose an automated data construction pipeline that generates paired real-fake videos by replacing temporal segments with boundary-frame-conditioned video generation models. Furthermore, we introduce \textbf{Evidence-Guided Reward Redistribution}, which performs evidence-aware credit assignment by redistributing rewards among label-correct responses according to evidence quality. This preserves reliable label supervision while encouraging detectors to acquire fine-grained and verifiable forgery localization capabilities. Extensive experiments demonstrate that \textbf{VidForensics-M1} effectively leverages verifiable temporal evidence to achieve robust and generalizable AI-generated video detection.

cs.CV↗

arXiv2026

OmniVerifier-M1: Multimodal Meta-Verifier with Explicit Structured Recalibration

Visual outcomes are increasingly central to multimodal large language models, making reliable and fine-grained verification essential for scaling generalist foundation models. In this work, we investigate multimodal meta-verification, which leverages verifier-generated rationales rather than decision-only signals, and explore how to effectively incorporate meta-verification feedback into multimodal verifier training. We identify two key findings. First, symbolic verifier outputs (e.g., bounding boxes) outperform textual explanations as meta-verification rationales, enabling efficient rule-based reinforcement learning rewards while avoiding reliance on model-based rewards from auxiliary judge models. Second, decoupling reinforcement learning objectives for binary judgment and meta-verification substantially outperforms joint reward optimization, due to intrinsic differences in output structure and learning dynamics. Based on these insights, we train OmniVerifier-M1, a generalist visual verifier leveraging symbolic meta-verification and decoupled reinforcement learning. OmniVerifier-M1 provides robust verification and fine-grained error localization, and further enables M1-TTS, a verifier-driven agentic generation system achieving dynamic region-level self-correction. This approach paves the way for more reliable, interpretable, and fine-grained multimodal verification, supporting safer and more controllable foundation model deployment.

cs.CL↗

arXiv2025

Calibrating Biased Distribution in VFM-derived Latent Space via Cross-Domain Geometric Consistency

Despite the fast progress of deep learning, one standing challenge is the gap of the observed training samples and the underlying true distribution. There are multiple reasons for the causing of this gap e.g. sampling bias, noise etc. In the era of foundation models, we show that when leveraging the off-the-shelf (vision) foundation models (e.g., CLIP, DINOv2) for feature extraction, the geometric shapes of the resulting feature distributions exhibit remarkable transferability across domains and datasets. To verify its practical usefulness, we embody our geometric knowledge-guided distribution calibration framework in two popular and challenging settings: federated learning and long-tailed recognition. In the federated setting, we devise a technique of acquiring the global geometric shape under privacy constraints, then leverage this knowledge to generate new samples for clients, in the aim of bridging the gap between local and global observations. In long-tailed learning, it utilizes the geometric knowledge transferred from sample-rich categories to recover the true distribution for sample-scarce tail classes. Comprehensive experiments show that our proposed geometric knowledge-guided distribution calibration effectively overcomes information deficits caused by data heterogeneity and sample imbalance, with boosted performance across benchmarks.

cs.CV↗

arXiv2025

Compositional Attribute Imbalance in Vision Datasets

Visual attribute imbalance is a common yet underexplored issue in image classification, significantly impacting model performance and generalization. In this work, we first define the first-level and second-level attributes of images and then introduce a CLIP-based framework to construct a visual attribute dictionary, enabling automatic evaluation of image attributes. By systematically analyzing both single-attribute imbalance and compositional attribute imbalance, we reveal how the rarity of attributes affects model performance. To tackle these challenges, we propose adjusting the sampling probability of samples based on the rarity of their compositional attributes. This strategy is further integrated with various data augmentation techniques (such as CutMix, Fmix, and SaliencyMix) to enhance the model's ability to represent rare attributes. Extensive experiments on benchmark datasets demonstrate that our method effectively mitigates attribute imbalance, thereby improving the robustness and fairness of deep neural networks. Our research highlights the importance of modeling visual attribute distributions and provides a scalable solution for long-tail image classification tasks.

cs.CV↗

arXiv2025

ProTCT: Projection quantification and fidelity constraint integrated deep reconstruction for Tangential CT

Tangential computed tomography (TCT) is a useful tool for imaging the large-diameter samples, such as oil pipelines and rockets. However, TCT projections are truncated along the detector direction, resulting in degraded slices with radial artifacts. Meanwhile, existing methods fail to reconstruct decent images because of the ill-defined sampling condition in the projection domain and oversmoothing in the cross-section domain. In this paper, we propose a projection quantification and fidelity constraint integrated deep TCT reconstruction method (ProTCT) to improve the slice quality. Specifically, the sampling conditions for reconstruction are analysed, offering practical guidelines for TCT system design. Besides, a deep artifact-suppression network together with a fidelity-constraint module that operates across both projection and cross-section domains to remove artifacts and restore edge details. Demonstrated on simulated and real datasets, the ProTCT shows good performance in structure restoration and detail retention. This work contributes to exploring the sampling condition and improving the slice quality of TCT, further promoting the application of large view field CT imaging.

eess.IV↗

arXiv2025

Geometric Origins of Bias in Deep Neural Networks: A Human Visual System Perspective

Bias formation in deep neural networks (DNNs) remains a critical yet poorly understood challenge, influencing both fairness and reliability in artificial intelligence systems. Inspired by the human visual system, which decouples object manifolds through hierarchical processing to achieve object recognition, we propose a geometric analysis framework linking the geometric complexity of class-specific perceptual manifolds in DNNs to model bias. Our findings reveal that differences in geometric complexity can lead to varying recognition capabilities across categories, introducing biases. To support this analysis, we present the Perceptual-Manifold-Geometry library, designed for calculating the geometric properties of perceptual manifolds. The toolkit has been downloaded and installed over 4,500 times. This work provides a novel geometric perspective on bias formation in modern learning systems and lays a theoretical foundation for developing more equitable and robust artificial intelligence.

cs.CV↗

arXiv2024

Multi wavefunction overlap and multi entropy for topological ground states in (2+1) dimensions

Multi-wavefunction overlaps -- generalizations of the quantum mechanical inner product for more than two quantum many-body states -- are valuable tools for studying many-body physics. In this paper, we investigate the multi-wavefunction overlap of (2+1)-dimensional gapped ground states, focusing particularly on symmetry-protected topological (SPT) states. We demonstrate how these overlaps can be calculated using the bulk-boundary correspondence and (1+1)-dimensional edge theories, specifically conformal field theory. When applied to SPT phases, we show that the topological invariants, which can be thought of as discrete higher Berry phases, can be extracted from the multi-wavefunction overlap of four ground states with appropriate symmetry actions. Additionally, we find that the multi-wavefunction overlap can be expressed in terms of the realignment of reduced density matrices. Furthermore, we illustrate that the same technique can be used to evaluate the multi-entropy -- a quantum information theoretical quantity associated with multi-partition of many-body quantum states -- for (2+1)-dimensional gapped ground states. Combined with numerics, we show that the difference between multi-entropy for tripartition and second R\'enyi entropies is bounded from below by $(c_{{\it tot}}/4)\ln 2$ where $c_{{\it tot}}$ is the central charge of ungappable degrees of freedom. To calculate multi-entropy numerically for free fermion systems (such as Chern insulators), we develop the correlator method for multi-entropy.

cond-mat.str-el↗

arXiv2023

MyoFold: rapid Myocardial tissue and movement quantification via a highly Folded sequence

Purpose: To develop and evaluate a cardiovascular magnetic resonance sequence (MyoFold) for rapid myocardial tissue and movement characterization. Method: MyoFold sequentially performs joint T1/T2 mapping and cine for one left-ventricle slice within a breathing-holding of 12 heartbeats. MyoFold uses balanced Steady-State-Free-Precession (bSSFP) with 2-fold acceleration for data readout and adopts an electrocardiogram (ECG) to synchronize the cardiac cycle. MyoFold first acquires six single-shot inversion-recovery images at the diastole of the first six heartbeats. For joint T1/T2 mapping, T2 preparation (T2-prep) adds different T2 weightings to the last three images. On the remaining six heartbeats, segmented bSSFP is continuously performed for each cardiac phase for cine. We build a neural network and trained it using the numerical simulation of MyoFold for T1 and T2 calculations. MyoFold was validated through phantom and in-vivo experiments and compared to MOLLI, SASHA, T2-prep bSSFP, and convention cine. Results: MyoFold phantom T1 had a 10% overestimation while MyoFold T2 had high accuracy. MyoFold in-vivo T1 had comparable accuracy to that of SASHA and precision to that of MOLLI. MyoFold had good agreement with T2-prep bSSFP in myocardium T2 measurement. There was no difference in the myocardium thickness measurement between the MyoFold cine and convention cine. Conclusion: MyoFold can simultaneously quantify myocardial tissue and movement, with accuracy and precision comparable to dedicated sequences, saving three-fold scan time.

physics.med-ph↗

arXiv2023

Notes on a Surface Defect in the $O(N)$ Model

We study a surface defect in the free and critical $O(N)$ vector models, defined by adding a quadratic perturbation localized on a two-dimensional subspace of the $d$-dimensional CFT. We compute the beta function for the corresponding defect renormalization group (RG) flow, and provide evidence that at long distances the system flows to a nontrivial defect conformal field theory (DCFT). We use epsilon and large $N$ expansions to compute several physical quantities in the DCFT, finding agreement across different expansion methods. We also compute the defect free energy, and check consistency with the so-called $b$-theorem for RG flows on surface defects.

hep-th↗

arXiv2022

Entanglement Entropy of Free Fermions in Timelike Slices

We define the entanglement entropy of free fermion quantum states in an arbitrary spacetime slice of a discrete set of points, and particularly investigate timelike (causal) slices. For 1D lattice free fermions with an energy bandwidth $E_0$, we calculate the time-direction entanglement entropy $S_A$ in a time-direction slice of a set of times $t_n=n\tau$ ($1\le n\le K$) spanning a time length $t$ on the same site. For zero temperature ground states, we find that $S_A$ shows volume law when $\tau\gg\tau_0=2\pi/E_0$; in contrast, $S_A\sim \frac{1}{3}\ln t$ when $\tau=\tau_0$, and $S_A\sim\frac{1}{6}\ln t$ when $\tau<\tau_0$, resembling the Calabrese-Cardy formula for one flavor of nonchiral and chiral fermion, respectively. For finite temperature thermal states, the mutual information also saturates when $\tau<\tau_0$. For non-eigenstates, volume law in $t$ and signatures of the Lieb-Robinson bound velocity can be observed in $S_A$. For generic spacetime slices with one point per site, the zero temperature entanglement entropy shows a clear transition from area law to volume law when the slice varies from spacelike to timelike.

cond-mat.stat-mech↗

arXiv2022

Koopman Methods for Estimation of Animal Motions over Unknown Submanifolds

This paper introduces a data-dependent approximation of the forward kinematics map for certain types of animal motion models. It is assumed that motions are supported on a low-dimensional, unknown configuration manifold $Q$ that is regularly embedded in high dimensional Euclidean space $X:=\mathbb{R}^d$. This paper introduces a method to estimate forward kinematics from the unknown configuration submanifold $Q$ to an $n$-dimensional Euclidean space $Y:=\mathbb{R}^n$ of observations. A known reproducing kernel Hilbert space (RKHS) is defined over the ambient space $X$ in terms of a known kernel function, and computations are performed using the known kernel defined on the ambient space $X$. Estimates are constructed using a certain data-dependent approximation of the Koopman operator defined in terms of the known kernel on $X$. However, the rate of convergence of approximations is studied in the space of restrictions to the unknown manifold $Q$. Strong rates of convergence are derived in terms of the fill distance of samples in the unknown configuration manifold, provided that a novel regularity result holds for the Koopman operator. Additionally, we show that the derived rates of convergence can be applied in some cases to estimates generated by the extended dynamic mode decomposition (EDMD) method. We illustrate characteristics of the estimates for simulated data as well as samples collected during motion capture experiments.

stat.ML↗

arXiv2021

Revisiting Point Cloud Shape Classification with a Simple and Effective Baseline

Processing point cloud data is an important component of many real-world systems. As such, a wide variety of point-based approaches have been proposed, reporting steady benchmark improvements over time. We study the key ingredients of this progress and uncover two critical results. First, we find that auxiliary factors like different evaluation schemes, data augmentation strategies, and loss functions, which are independent of the model architecture, make a large difference in performance. The differences are large enough that they obscure the effect of architecture. When these factors are controlled for, PointNet++, a relatively older network, performs competitively with recent methods. Second, a very simple projection-based method, which we refer to as SimpleView, performs surprisingly well. It achieves on par or better results than sophisticated state-of-the-art methods on ModelNet40 while being half the size of PointNet++. It also outperforms state-of-the-art methods on ScanObjectNN, a real-world point cloud benchmark, and demonstrates better cross-dataset generalization. Code is available at https://github.com/princeton-vl/SimpleView.

cs.CV↗

arXiv2018

Learning Hash Codes via Hamming Distance Targets

We present a powerful new loss function and training scheme for learning binary hash codes with any differentiable model and similarity function. Our loss function improves over prior methods by using log likelihood loss on top of an accurate approximation for the probability that two inputs fall within a Hamming distance target. Our novel training scheme obtains a good estimate of the true gradient by better sampling inputs and evaluating loss terms between all pairs of inputs in each minibatch. To fully leverage the resulting hashes, we use multi-indexing. We demonstrate that these techniques provide large improvements to a similarity search tasks. We report the best results to date on competitive information retrieval tasks for ImageNet and SIFT 1M, improving MAP from 73% to 84% and reducing query cost by a factor of 2-8, respectively.

cs.LG↗

arXiv2018

Convolutional Hashing for Automated Scene Matching

We present a powerful new loss function and training scheme for learning binary hash functions. In particular, we demonstrate our method by creating for the first time a neural network that outperforms state-of-the-art Haar wavelets and color layout descriptors at the task of automated scene matching. By accurately relating distance on the manifold of network outputs to distance in Hamming space, we achieve a 100-fold reduction in nontrivial false positive rate and significantly higher true positive rate. We expect our insights to provide large wins for hashing models applied to other information retrieval hashing tasks as well.

cs.CV↗

arXiv2016

Differential Harnack Estimates for Fisher's Equation

In this paper, we derive several differential Harnack estimates (also known as Li-Yau-Hamilton-type estimates) for positive solutions of Fisher's equation. We use the estimates to obtain lower bounds on the speed of traveling wave solutions and to construct classical Harnack inequalities.

math.AP↗