arXiv subjects

Xincheng Yao

Publications and source records attributed to Xincheng Yao.

At least 19 recordsLinked to original sources

arXiv2026

Back to the Definition: Estimating Step-Level Advantages via Trajectory Graphs for Agentic Reinforcement Learning

Group-based reinforcement learning (RL) methods, such as GRPO and its variants, have become a leading paradigm for training reasoning and agentic large language models (LLMs). While their group-normalized advantage estimation is reliable at the response level, it becomes systematically biased at the step level, since coarse-grained trajectory-level advantages are hard to accurately reflect the contribution of individual steps (i.e, failed trajectories may contain valuable steps). Revisiting the foundational RL definition, we notice that GRPO's success on single-turn tasks stems from its advantage estimation strategy, which adheres to the basic definition: the mean reward of multiple actions sampled from the same state constitutes a credible state-value estimate. Extending the faithful estimation to step-level would in principle demand sampling multiple actions from each intermediate state, which is too costly on a per-state basis. To mitigate this issue, we propose a Graph-based Faithful sTep-level credit-assignment framework (GRAFT) that grafts all rollout trajectories into a trajectory graph, recovering node state-values via Bellman iteration on the graph, and assigning credit to each edge by the node value difference. Theoretically, the estimated step-level advantage faithfully adheres to the basic advantage definition in RL. To further ensure the reliability of step-level advantage estimation, we further propose Graph GAE, which extends GAE to the trajectory graph for reducing the impact of state-value estimation bias. Experiments across a range of multi-turn agentic benchmarks show consistent gains over GRPO and superior performance compared to recent agentic RL algorithms. Code will be available at https://github.com/xcyao00/GRAFT.

cs.AI↗

arXiv2026

HTPO: Towards Exploration-Exploitation Balanced Policy Optimization via Hierarchical Token-level Objective Control

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a pivotal technique for enhancing the reasoning capabilities of Large Language Models (LLMs). However, the de facto practice of mainstream RL algorithms is to treat all tokens of one response equally and assign the same optimization objective to each token, failing to provide granular guidance for the reasoning process. While in Chain-of-Thought (CoT) reasoning, different tokens usually play distinct roles. Therefore, the current RL algorithms lack an effective mechanism to dynamically balance the exploration-exploitation trade-off during learning. To this end, we propose Hierarchical Token-level Objective Control Policy Optimization (HTPO), a novel RL algorithm that takes the divide-and-conquer idea to hierarchically partition the response tokens into specific functional groups from three aspects (i.e., prompt difficulty, answer correctness, and token entropy). Within each group, according to the contributions to exploration or exploitation, we design specialized optimization objectives to facilitate the effective execution of each token's expected functionality. In this way, HTPO can achieve a more balanced exploration-exploitation trade-off. Extensive experiments on challenging reasoning benchmarks validate the superiority of our HTPO algorithm, which significantly outperforms the strong DAPO baseline (e.g., +8.6% and +6.7% on AIME'24 and AIME'25, respectively). When scaling test-time compute, the HTPO-trained model maintains a consistent performance advantage over the DAPO baseline, and the gap widens as the sampling budget increases, validating that our adaptive token-level control method fosters effective exploration without sacrificing exploitation performance. Code will be at https://github.com/xcyao00/HTPO.

cs.LG↗

arXiv2026

MMR-AD: A Large-Scale Multimodal Dataset for Benchmarking General Anomaly Detection with Multimodal Large Language Models

In the progress of industrial anomaly detection, general anomaly detection (GAD) is an emerging trend and also the ultimate goal. Unlike the conventional single- and multi-class AD, general AD aims to train a general AD model that can directly detect anomalies in diverse novel classes without any retraining or fine-tuning on the target data. Recently, Multimodal Large Language Models (MLLMs) have shown great promise in achieving general anomaly detection due to their revolutionary visual understanding and language reasoning capabilities. However, MLLM's general AD ability remains underexplored due to: (1) MLLMs are pretrained on amounts of data sourced from the Web, these data still have significant gaps with the data in AD scenarios. Moreover, the image-text pairs during pretraining are also not specifically for AD tasks. (2) The current mainstream AD datasets are image-based and not yet suitable for post-training MLLMs. To facilitate MLLM-based general AD research, we present MMR-AD, which is a comprehensive benchmark for both training and evaluating MLLM-based AD models. With MMR-AD, we reveal that the AD performance of current SOTA generalist MLLMs still falls far behind the industrial requirements. Based on MMR-AD, we also propose a baseline model, Anomaly-R1, which is a reasoning-based AD model that learns from the CoT data in MMR-AD and is further enhanced by reinforcement learning. Extensive experiments show that our Anomaly-R1 achieves remarkable improvements over generalist MLLMs in both anomaly detection and localization.

cs.CV↗

arXiv2026

Self-locking non-volatile coding metasurfaces via origami-based mechanical bits

Digital coding metasurfaces have revolutionized electromagnetic (EM) manipulation, yet typical tunable approaches based on active components suffer from the "volatility" bottleneck. While mechanical modulation provides a potential solution, current implementations generally lack inherent state-locking capability, rendering them vulnerable to environmental disturbances and actuation errors. Inspired by the concept of mechanical bits (MBs), this paper presents a self-locking non-volatile coding metasurface platform enabled by Kresling origami-based MBs, where the continuous mechanical deformation of individual meta-atoms is discretized into robust binary geometric states protected by intrinsic energy barriers. The bistable states are strictly mapped to 1-bit EM coding phases via tailored metallic patterns integrated onto a multimaterial 3D printed Kresling origami array. Building upon this concept, both transmission- and reflection-type prototypes are proposed and experimentally demonstrated, exhibiting exceptional wavefront manipulation capabilities through near-field holographic imaging and far-field beam steering. In addition, the lightweight origami unit (1.5 g) exhibits an exceptional load-bearing capacity, supporting over 100 times its own weight. These results bridge mechanical logic with EM information processing, establishing a universal physical paradigm for constructing low-power, highly robust coding metasurfaces resilient to extreme environments.

physics.optics↗

arXiv2025

ADPretrain: Advancing Industrial Anomaly Detection via Anomaly Representation Pretraining

The current mainstream and state-of-the-art anomaly detection (AD) methods are substantially established on pretrained feature networks yielded by ImageNet pretraining. However, regardless of supervised or self-supervised pretraining, the pretraining process on ImageNet does not match the goal of anomaly detection (i.e., pretraining in natural images doesn't aim to distinguish between normal and abnormal). Moreover, natural images and industrial image data in AD scenarios typically have the distribution shift. The two issues can cause ImageNet-pretrained features to be suboptimal for AD tasks. To further promote the development of the AD field, pretrained representations specially for AD tasks are eager and very valuable. To this end, we propose a novel AD representation learning framework specially designed for learning robust and discriminative pretrained representations for industrial anomaly detection. Specifically, closely surrounding the goal of anomaly detection (i.e., focus on discrepancies between normals and anomalies), we propose angle- and norm-oriented contrastive losses to maximize the angle size and norm difference between normal and abnormal features simultaneously. To avoid the distribution shift from natural images to AD images, our pretraining is performed on a large-scale AD dataset, RealIAD. To further alleviate the potential shift between pretraining data and downstream AD datasets, we learn the pretrained AD representations based on the class-generalizable representation, residual features. For evaluation, based on five embedding-based AD methods, we simply replace their original features with our pretrained representations. Extensive experiments on five AD datasets and five backbones consistently show the superiority of our pretrained features. The code is available at https://github.com/xcyao00/ADPretrain.

cs.CV↗

arXiv2025

ResAD++: Towards Class Agnostic Anomaly Detection via Residual Feature Learning

This paper explores the problem of class-agnostic anomaly detection (AD), where the objective is to train one class-agnostic AD model that can generalize to detect anomalies in diverse new classes from different domains without any retraining or fine-tuning on the target data. When applied for new classes, the performance of current single- and multi-class AD methods is still unsatisfactory. One fundamental reason is that representation learning in existing methods is still class-related, namely, feature correlation. To address this issue, we propose residual features and construct a simple but effective framework, termed ResAD. Our core insight is to learn the residual feature distribution rather than the initial feature distribution. Residual features are formed by matching and then subtracting normal reference features. In this way, we can effectively realize feature decorrelation. Even in new classes, the distribution of normal residual features would not remarkably shift from the learned distribution. In addition, we think that residual features still have one issue: scale correlation. To this end, we propose a feature hypersphere constraining approach, which learns to constrain initial normal residual features into a spatial hypersphere for enabling the feature scales of different classes as consistent as possible. Furthermore, we propose a novel logbarrier bidirectional contraction OCC loss and vector quantization based feature distribution matching module to enhance ResAD, leading to the improved version of ResAD (ResAD++). Comprehensive experiments on eight real-world AD datasets demonstrate that our ResAD++ can achieve remarkable AD results when directly used in new classes, outperforming state-of-the-art competing methods and also surpassing ResAD. The code is available at https://github.com/xcyao00/ResAD.

cs.CV↗

arXiv2025

Beyond Label Semantics: Language-Guided Action Anatomy for Few-shot Action Recognition

Few-shot action recognition (FSAR) aims to classify human actions in videos with only a small number of labeled samples per category. The scarcity of training data has driven recent efforts to incorporate additional modalities, particularly text. However, the subtle variations in human posture, motion dynamics, and the object interactions that occur during different phases, are critical inherent knowledge of actions that cannot be fully exploited by action labels alone. In this work, we propose Language-Guided Action Anatomy (LGA), a novel framework that goes beyond label semantics by leveraging Large Language Models (LLMs) to dissect the essential representational characteristics hidden beneath action labels. Guided by the prior knowledge encoded in LLM, LGA effectively captures rich spatiotemporal cues in few-shot scenarios. Specifically, for text, we prompt an off-the-shelf LLM to anatomize labels into sequences of atomic action descriptions, focusing on the three core elements of action (subject, motion, object). For videos, a Visual Anatomy Module segments actions into atomic video phases to capture the sequential structure of actions. A fine-grained fusion strategy then integrates textual and visual features at the atomic level, resulting in more generalizable prototypes. Finally, we introduce a Multimodal Matching mechanism, comprising both video-video and video-text matching, to ensure robust few-shot classification. Experimental results demonstrate that LGA achieves state-of-the-art performance across multipe FSAR benchmarks.

cs.CV↗

arXiv2025

HRVVS: A High-resolution Video Vasculature Segmentation Network via Hierarchical Autoregressive Residual Priors

The segmentation of the hepatic vasculature in surgical videos holds substantial clinical significance in the context of hepatectomy procedures. However, owing to the dearth of an appropriate dataset and the inherently complex task characteristics, few researches have been reported in this domain. To address this issue, we first introduce a high quality frame-by-frame annotated hepatic vasculature dataset containing 35 long hepatectomy videos and 11442 high-resolution frames. On this basis, we propose a novel high-resolution video vasculature segmentation network, dubbed as HRVVS. We innovatively embed a pretrained visual autoregressive modeling (VAR) model into different layers of the hierarchical encoder as prior information to reduce the information degradation generated during the downsampling process. In addition, we designed a dynamic memory decoder on a multi-view segmentation network to minimize the transmission of redundant information while preserving more details between frames. Extensive experiments on surgical video datasets demonstrate that our proposed HRVVS significantly outperforms the state-of-the-art methods. The source code and dataset will be publicly available at \{https://github.com/scott-yjyang/HRVVS}.

cs.CV↗

arXiv2024

ResAD: A Simple Framework for Class Generalizable Anomaly Detection

This paper explores the problem of class-generalizable anomaly detection, where the objective is to train one unified AD model that can generalize to detect anomalies in diverse classes from different domains without any retraining or fine-tuning on the target data. Because normal feature representations vary significantly across classes, this will cause the widely studied one-for-one AD models to be poorly classgeneralizable (i.e., performance drops dramatically when used for new classes). In this work, we propose a simple but effective framework (called ResAD) that can be directly applied to detect anomalies in new classes. Our main insight is to learn the residual feature distribution rather than the initial feature distribution. In this way, we can significantly reduce feature variations. Even in new classes, the distribution of normal residual features would not remarkably shift from the learned distribution. Therefore, the learned model can be directly adapted to new classes. ResAD consists of three components: (1) a Feature Converter that converts initial features into residual features; (2) a simple and shallow Feature Constraintor that constrains normal residual features into a spatial hypersphere for further reducing feature variations and maintaining consistency in feature scales among different classes; (3) a Feature Distribution Estimator that estimates the normal residual feature distribution, anomalies can be recognized as out-of-distribution. Despite the simplicity, ResAD can achieve remarkable anomaly detection results when directly used in new classes. The code is available at https://github.com/xcyao00/ResAD.

cs.CV↗

arXiv2024

Hierarchical Gaussian Mixture Normalizing Flow Modeling for Unified Anomaly Detection

Unified anomaly detection (AD) is one of the most challenges for anomaly detection, where one unified model is trained with normal samples from multiple classes with the objective to detect anomalies in these classes. For such a challenging task, popular normalizing flow (NF) based AD methods may fall into a "homogeneous mapping" issue,where the NF-based AD models are biased to generate similar latent representations for both normal and abnormal features, and thereby lead to a high missing rate of anomalies. In this paper, we propose a novel Hierarchical Gaussian mixture normalizing flow modeling method for accomplishing unified Anomaly Detection, which we call HGAD. Our HGAD consists of two key components: inter-class Gaussian mixture modeling and intra-class mixed class centers learning. Compared to the previous NF-based AD methods, the hierarchical Gaussian mixture modeling approach can bring stronger representation capability to the latent space of normalizing flows, so that even complex multi-class distribution can be well represented and learned in the latent space. In this way, we can avoid mapping different class distributions into the same single Gaussian prior, thus effectively avoiding or mitigating the "homogeneous mapping" issue. We further indicate that the more distinguishable different class centers, the more conducive to avoiding the bias issue. Thus, we further propose a mutual information maximization loss for better structuring the latent feature space. We evaluate our method on four real-world AD benchmarks, where we can significantly improve the previous NF-based AD methods and also outperform the SOTA unified AD methods.

cs.LG↗

arXiv2023

Focus the Discrepancy: Intra- and Inter-Correlation Learning for Image Anomaly Detection

Humans recognize anomalies through two aspects: larger patch-wise representation discrepancies and weaker patch-to-normal-patch correlations. However, the previous AD methods didn't sufficiently combine the two complementary aspects to design AD models. To this end, we find that Transformer can ideally satisfy the two aspects as its great power in the unified modeling of patch-wise representations and patch-to-patch correlations. In this paper, we propose a novel AD framework: FOcus-the-Discrepancy (FOD), which can simultaneously spot the patch-wise, intra- and inter-discrepancies of anomalies. The major characteristic of our method is that we renovate the self-attention maps in transformers to Intra-Inter-Correlation (I2Correlation). The I2Correlation contains a two-branch structure to first explicitly establish intra- and inter-image correlations, and then fuses the features of two-branch to spotlight the abnormal patterns. To learn the intra- and inter-correlations adaptively, we propose the RBF-kernel-based target-correlations as learning targets for self-supervised learning. Besides, we introduce an entropy constraint strategy to solve the mode collapse issue in optimization and further amplify the normal-abnormal distinguishability. Extensive experiments on three unsupervised real-world AD benchmarks show the superior performance of our approach. Code will be available at https://github.com/xcyao00/FOD.

cs.CV↗

arXiv2023

Explicit Boundary Guided Semi-Push-Pull Contrastive Learning for Supervised Anomaly Detection

Most anomaly detection (AD) models are learned using only normal samples in an unsupervised way, which may result in ambiguous decision boundary and insufficient discriminability. In fact, a few anomaly samples are often available in real-world applications, the valuable knowledge of known anomalies should also be effectively exploited. However, utilizing a few known anomalies during training may cause another issue that the model may be biased by those known anomalies and fail to generalize to unseen anomalies. In this paper, we tackle supervised anomaly detection, i.e., we learn AD models using a few available anomalies with the objective to detect both the seen and unseen anomalies. We propose a novel explicit boundary guided semi-push-pull contrastive learning mechanism, which can enhance model's discriminability while mitigating the bias issue. Our approach is based on two core designs: First, we find an explicit and compact separating boundary as the guidance for further feature learning. As the boundary only relies on the normal feature distribution, the bias problem caused by a few known anomalies can be alleviated. Second, a boundary guided semi-push-pull loss is developed to only pull the normal features together while pushing the abnormal features apart from the separating boundary beyond a certain margin region. In this way, our model can form a more explicit and discriminative decision boundary to distinguish known and also unseen anomalies from normal samples more effectively. Code will be available at https://github.com/xcyao00/BGAD.

cs.CV↗

arXiv2022

Normalized Blood Flow Index in Optical Coherence Tomography Angiography Provides a Sensitive Biomarker of Early Diabetic Retinopathy

Purpose: To evaluate the sensitivity of normalized blood flow index (NBFI) for detecting early diabetic retinopathy (DR). Methods: Optical coherence tomography angiography (OCTA) images of 30 eyes from 20 healthy controls, 21 eyes of diabetic patients with no DR (NoDR) and 26 eyes from 22 patients with mild non-proliferative DR (NPDR) were analyzed in this study. The OCTA images were centered on the fovea and covered a 6 mm x 6 mm area. Enface projections of the superficial vascular plexus (SVP) and the deep capillary plexus (DCP) were obtained for the quantitative OCTA feature analysis. Three quantitative OCTA features were examined: blood vessel density (BVD), blood flow flux (BFF), and normalized blood flow index (NBFI). Each feature was calculated from both the SVP and DCP and their sensitivity to distinguish the three cohorts of the study were evaluated. Results: The only quantitative feature that was capable of distinguishing between all three cohorts was NBFI in the DCP image. Comparative study revealed that both BVD and BFF were able to distinguish the controls from NoDR and mild NPDR. However, neither BVD nor BFF was sensitive enough to separate NoDR from the healthy controls. Conclusion: The NBFI has been demonstrated as a sensitive biomarker of early DR, revealing retinal blood flow abnormality better than traditional BVD and BFF. The NBFI in the DCP was verified as the most sensitive biomarker, supporting that diabetes affects the DCP earlier than SVP in DR.

physics.med-ph↗

arXiv2022

Intrinsic signal optoretinography of dark adaptation abnormality due to rod degeneration

Significance: Multiple eye diseases such as age-related macular degeneration, diabetic retinopathy, and retinitis pigmentosa can cause photoreceptor dysfunction. Rod photoreceptors are known to be more vulnerable than cone photoreceptors. Therefore, functional assessment of rod photoreceptors is important for early detection of eye diseases. Aim: This study is to demonstrate the feasibility of using intrinsic optical signal (IOS) optoretinography (ORG) for objective detection of dark adaptation (DA) abnormality due to rod photoreceptor degeneration. Approach: Functional optical coherence tomography (OCT) was employed for IOS ORG of wild-type (WT) and retinal degeneration 10 (rd10) mice. Six WT C57BL/6J and eight rd10 B6.CXB1-Pde6brd10/J mice at postnatal day 14 (P14) were used for this study. Dynamic OCT analysis of retinal thickness and brightness changes, corresponding to light to dark transition, was implemented Results: Comparative measurement of the retina under light and dark conditions revealed significant IOS changes within the outer retina. The thickness between external limiting membrane (ELM) and retinal pigment epithelium (RPE) reduced, and the OCT brightness of inner segment ellipsoid zone (EZ) decreased during DA, compared to light adaptation (LA). Relative EZ intensity change was observed to decrease larger in rd10, compared to WT. The relative intensity of the hyporeflective band between ELM and RPE also showed significant decrease in rd10 retinas during DA. Conclusions: DA-IOS abnormalities were observed in rd10, compared to WT at P14. The ORG measurement of DA-IOS kinetics promises valuable information for noninvasive assessment of photoreceptor function.

physics.bio-ph↗

arXiv2022

A portable widefield fundus camera with high dynamic range imaging capability

Fundus photography is indispensable for clinical detection and management of eye diseases. Limited image contrast and field of view (FOV) are common limitations of conventional fundus cameras, making it difficult to detect subtle abnormalities at the early stages of eye diseases. Further improvements of image contrast and FOV coverage are important to improve early disease detection and reliable treatment assessment. We report here a portable fundus camera, with a wide FOV and high dynamic range (HDR) imaging capabilities. Miniaturized indirect ophthalmoscopy illumination was employed to achieve the portable design for nonmydriatic, widefield fundus photography. Orthogonal polarization control was used to eliminate illumination reflectance artifact. With independent power controls, three fundus images were sequentially acquired and fused to achieve HDR function for local image contrast enhancement. A 101° eye-angle (67° visual-angle) snapshot FOV was achieved for nonmydriatic fundus photography. The effective FOV can be readily expanded up to 190° eye-angle (134° visual-angle) with the aid of a fixation target, without the need of pharmacologic pupillary dilation. The effectiveness of HDR imaging was validated with both normal healthy and pathologic eyes, compared to a conventional fundus camera.

physics.med-ph↗

arXiv2022

Functional Optical Coherence Tomography for Intrinsic Signal Optoretinography: Recent Developments and Deployment Challenges

Intrinsic optical signal (IOS) imaging of the retina, also termed as optoretinography (ORG), promises a noninvasive method for objective assessment of retinal function. By providing unparalleled capability to differentiate individual layers of the retina, functional optical coherence tomography (OCT) has been actively investigated for intrinsic signal ORG measurements. However, clinical deployment of functional OCT for quantitative ORG is still challenging due to the lack of a standardized imaging protocol and the complication of IOS sources and mechanisms. This article aims to summarize recent developments of functional OCT for ORG measurement, OCT intensity- and phase-based IOS processing. Technical challenges and perspectives of quantitative IOS analysis and ORG interpretations are discussed.

physics.med-ph↗

arXiv2022

Depth-resolved vascular profile features for artery-vein classification in OCT and OCT angiography of human retina

This study is to characterize reflectance profiles of retinal blood vessels in optical coherence tomography (OCT), and to validate these vascular features to guide artery-vein classification in OCT angiography (OCTA) of human retina. Depth-resolved OCT reveals unique features of retinal arteries and veins. Retinal arteries show hyper-reflective boundaries at both upper (inner side towards the vitreous) and lower (outer side towards the choroid) walls. In contrary, retinal veins reveal hyper-reflectivity at the upper boundary only. Uniform lumen intensity was observed in both small and large arteries. However, the vein lumen intensity was dependent on the vessel size. Small veins exhibit a hyper-reflective zone at the bottom half of the lumen, while large veins show a hypo-reflective zone at the bottom half of the lumen

q-bio.TO↗

arXiv2022

ADC-Net: An Open-Source Deep Learning Network for Automated Dispersion Compensation in Optical Coherence Tomography

Chromatic dispersion is a common problem to degrade the system resolution in optical coherence tomography (OCT). This study is to develop a deep learning network for automated dispersion compensation (ADC-Net) in OCT. The ADC-Net is based on a redesigned UNet architecture which employs an encoder-decoder pipeline. The input section encompasses partially compensated OCT B-scans with individual retinal layers optimized. Corresponding output is a fully compensated OCT B-scans with all retinal layers optimized. Two numeric parameters, i.e., peak signal to noise ratio (PSNR) and structural similarity index metric computed at multiple scales (MS-SSIM), were used for objective assessment of the ADC-Net performance. Comparative analysis of training models, including single, three, five, seven and nine input channels were implemented. The five-input channels implementation was observed as the optimal mode for ADC-Net training to achieve robust dispersion compensation in OCT

eess.IV↗