Author name cluster

Xin Cheng

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

15 papers

2 author rows

EAAI Journal 2026 Journal Article

Anti-forensic for quantization steps estimation based on direct and preemptive adversarial attacks

Jiawei Zhang
Mengjie Wang
Hao Wu
Xin Cheng
Xiangyang Luo
Bin Ma
Hao Wang
Jinwei Wang

Joint Photographic Experts Group (JPEG) quantization steps estimation aims to reveal the compressed history of the images, which can serve as an essential component and powerful technique to support forensics. Nowadays, various deep learning-based estimation methods have been proposed to achieve higher accuracy. However, due to supposing the ideal secure conditions of estimation, their robustness against deliberate attacks (especially adversarial attacks) has not been thoroughly studied, which poses a significant threat to their reliability. To address this issue, as the first attempt, we investigate the robustness of deep learning-based estimation methods against adversarial attacks, which can significantly deteriorate estimation accuracy without noticeable distortion. Specifically, we introduce a generation-based adversarial attack framework and propose two types of anti-forensic attacks, Direct Attack (DA) and Preemptive Attack (PA), to craft adversarial examples on double and single compressed images. To maximize the attack ability, we study the effect of regression and classification objectives on the adversarial property and design a joint loss function for stable and smooth optimization. Extensive experiments prove that the proposed DA and PA can achieve a high attack ability with low perturbation magnitude and satisfactory visual quality. More importantly, the generated adversarial examples present superior transferability across different estimation models and datasets, which proves the generality of the proposed method and also reveals the vulnerability of the existing deep learning-based estimation methods towards adversarial examples. Our code will be publicly available soon.

Details DOI

YNICL Journal 2025 Journal Article

A deep learning approach versus expert clinician panel in the classification of posterior circulation infarction

Leon S. Edwards
Milanka Visser
Cecilia Cappelen-Smith
Dennis Cordato
Andrew Bivard
Leonid Churilov
Christopher Blair
James Thomas

BACKGROUND: Posterior circulation infarction (POCI) is common. Imaging techniques such as non-contrast-CT (NCCT) and diffusion-weighted-magnetic-resonance-imaging commonly fail to detect hyperacute POCI. Studies suggest expert inspection of Computed Tomography Perfusion (CTP) improves diagnosis of POCI. In many settings, there is limited access to specialist expertise. Deep-learning has been successfully applied to automate imaging interpretation. This study aimed to develop and validate a deep-learning approach for the classification of POCI using CTP. METHODS: Data were analysed from 3541-patients from the International-stroke-perfusion-registry (INSPIRE). All patients with baseline multimodal-CT and follow-up imaging performed at 24-48 h were identified. A cohort of 541-patients was constructed on a 1:3 POCI-to -reference-ratio for model analysis. A 3D-Dense-Convolutional-Network (DenseNet) was trained to classify patients into POCI or non-POCI using CTP-deconvolved-maps. Six-stroke-experts also independently classified patients based upon stepwise access to multimodal CT (mCT) data. DenseNet results were compared against expert clinician results. Model and clinician performance was evaluated using area-under-the-receiver-operating-curve, sensitivity, specificity, accuracy and precision. Clinician agreement was measured with the Fleiss-Kappa-statistic. RESULTS: Best mean clinician diagnostic accuracy, sensitivity and agreement was demonstrated after review of all mCT data (AUC: 0.81, Sensitivity: 0.65, Fleiss-Kappa-statistic: 0.73). There was a spectrum of individual clinician results with an AUC-range of 0.73-0.86. Best DenseNet performance was recorded with an input combination of NCCT and delay-time maps. The DenseNet model was superior to the best mean clinician performance (AUC: 0.87) and was due to enhanced sensitivity (DenseNET: 0.77, Clinician: 0.65). The degree to which the DenseNet model outperformed each clinician ranged and was clinician specific (AUC improvement 0.01-0.14). CONCLUSION: Comprehensive review of CTP improves diagnostic performance and agreement amongst clinicians. A DenseNet model was superior to best mean clinician performance. The degree of improvement varied by specific clinician. Development of a clinician-DenseNet approach may improve inter-clinician agreement and diagnostic accuracy. This approach may alleviate limited specialist services in resource constrained settings.

Details DOI

ICLR Conference 2025 Conference Paper

AFlow: Automating Agentic Workflow Generation

Jiayi Zhang 0017
Jinyu Xiang
Zhaoyang Yu 0004
Fengwei Teng
Xionghui Chen
Jiaqi Chen
Mingchen Zhuge
Xin Cheng

Large language models (LLMs) have demonstrated remarkable potential in solving complex tasks across diverse domains, typically by employing agentic workflows that follow detailed instructions and operational sequences. However, constructing these workflows requires significant human effort, limiting scalability and generalizability. Recent research has sought to automate the generation and optimization of these workflows, but existing methods still rely on initial manual setup and fall short of achieving fully automated and effective workflow generation. To address this challenge, we reformulate workflow optimization as a search problem over code-represented workflows, where LLM-invoking nodes are connected by edges. We introduce AFLOW, an automated framework that efficiently explores this space using Monte Carlo Tree Search, iteratively refining workflows through code modification, tree-structured experience, and execution feedback. Empirical evaluations across six benchmark datasets demonstrate AFLOW's efficacy, yielding a 5.7% average improvement over state-of-the-art baselines. Furthermore, AFLOW enables smaller models to outperform GPT-4o on specific tasks at 4.55% of its inference cost in dollars. The code is available at https://github.com/FoundationAgents/AFlow.

Details

AAAI Conference 2025 Conference Paper

EyEar: Learning Audio Synchronized Human Gaze Trajectory Based on Physics-Informed Dynamics

Xiaochuan Liu
Xin Cheng
Yuchong Sun
Xiaoxue Wu
Ruihua Song
Hao Sun
Denghao Zhang

Imitating how humans move their gaze in a visual scene is a vital research problem for both visual understanding and psychology, kindling crucial applications such as building alive virtual characters. Previous studies aim to predict gaze trajectories when humans are free-viewing an image, searching for required targets, or looking for clues to answer questions in an image. While these tasks focus on visual-centric scenarios, humans move their gaze also along with audio signal inputs in more common scenarios. To fill this gap, we introduce a new task that predicts human gaze trajectories in a visual scene with synchronized audio inputs and provide a new dataset containing 20k gaze points from 8 subjects. To effectively integrate audio information and simulate the dynamic process of human gaze motion, we propose a novel learning framework called EyEar (Eye moving while Ear listening) based on physics-informed dynamics, which considers three key factors to predict gazes: eye inherent motion tendency, vision salient attraction, and audio semantic attraction. We also propose a probability density score to overcome the high individual variability of gaze trajectories, thereby improving the stabilization of optimization and the reliability of the evaluation. Experimental results show that EyEar outperforms all the baselines in the context of all evaluation metrics, thanks to the proposed components in the learning model.

PDF Details DOI

TMLR Journal 2025 Journal Article

Synthetic Data (Almost) from Scratch: Generalized Instruction Tuning for Language Models

Haoran Li
Qingxiu Dong
Zhengyang Tang
Chaojun Wang
Xingxing Zhang
Haoyang Huang
Shaohan Huang
Xiaolong Huang

We introduce Generalized Instruction Tuning (called GLAN), a general and scalable method for instruction tuning of Large Language Models (LLMs). Unlike prior work that relies on seed examples or existing datasets to construct instruction-tuning data, GLAN exclusively utilizes a pre-curated taxonomy of human knowledge and capabilities as input and generates large-scale synthetic instruction data across all disciplines. Specifically, inspired by the systematic structure in human education system, we build the taxonomy by decomposing human knowledge and capabilities to various fields, sub-fields and ultimately, distinct disciplines semi-automatically, facilitated by LLMs. Subsequently, we generate a comprehensive list of subjects for every discipline and proceed to design a syllabus tailored to each subject, again utilizing LLMs. With the fine-grained key concepts detailed in every class session of the syllabus, we are able to generate diverse instructions with a broad coverage across the entire spectrum of human knowledge and skills. Extensive experiments on large language models (e.g., Mistral) demonstrate that GLAN excels in multiple dimensions from mathematical reasoning, coding, academic exams, logical reasoning to general instruction following without using task-specific training data of these tasks. In addition, GLAN allows for easy customization and new fields or skills can be added by simply incorporating a new node into our taxonomy. While promising, our approach may inherit biases or inaccuracies from LLM-generated data as in other synthetic data work and is primarily evaluated on exam-style benchmarks. Broader evaluations and data quality control are left for future work.

PDF Details

YNIMG Journal 2025 Journal Article

UTE MRI for assessing demyelination in an mTBI mouse model: An open-field low-intensity blast study

Yajun Ma
Qingbo Tang
Xin Cheng
Jiyo S. Athertya
David Coughlin
Eric Y. Chang
Catherine E. Johnson
Jiankun Cui

Mild traumatic brain injury (mTBI) is a leading cause of long-term disability. Following mTBI, secondary chemical cascades and neuroinflammation can result in myelin damage, significantly impairing cognitive function. This study aims to assess demyelination in mice with mTBI induced by open-field low-intensity blast (LIB) using a novel three-dimensional short repetition time adiabatic inversion recovery UTE (3D STAIR-UTE) magnetic resonance imaging (MRI) sequence. Thirty male C57BL/6 mice, with 15 experiencing mTBI and 15 serving as sham controls, were included in this study. Behavioral tests were performed starting at 5 days post-injury to assess motor activity and anxiety-like responses followed by STAIR-UTE imaging using a pre-clinical 3T MRI scanner. Additionally, a proton density-weighted UTE sequence was scanned alongside the STAIR-UTE for quantification of myelin proton fraction (MPF). Luxol fast blue (LFB) staining was performed to evaluate myelin changes between the mTBI group and the control group. The behavioral tests indicated decreased motor activity in the center zone and increased anxiety-like response in the mTBI mice compared to sham controls. The STAIR-UTE sequence revealed significantly lower MPFs in the corpus callosum of mTBI mice (8.4 ± 0.4 % vs. 8.7 ± 0.4 %; P = 0.003), consistent with the myelin reduction observed in the LFB staining (0.77 ± 0.22 vs. 1.09 ± 0.15; P = 0.004). Our findings demonstrate that the STAIR-UTE sequence facilitates quantitative myelin imaging at 3T MRI, enabling the detection of demyelination in the white matter of the mouse brain associated with alterations in motor and anxiety domains post-LIB exposure.

Details DOI

NeurIPS Conference 2024 Conference Paper

xRAG: Extreme Context Compression for Retrieval-augmented Generation with One Token

Xin Cheng
Xun Wang
Xingxing Zhang
Tao Ge
Si-Qing Chen
Furu Wei
Huishuai Zhang
Dongyan Zhao

This paper introduces xRAG, an innovative context compression method tailored for retrieval-augmented generation. xRAG reinterprets document embeddings in dense retrieval--traditionally used solely for retrieval--as features from the retrieval modality. By employing a modality fusion methodology, xRAG seamlessly integrates these embeddings into the language model representation space, effectively eliminating the need for their textual counterparts and achieving an extreme compression rate. In xRAG, the only trainable component is the modality bridge, while both the retriever and the language model remain frozen. This design choice allows for the reuse of offline-constructed document embeddings and preserves the plug-and-play nature of retrieval augmentation. Experimental results demonstrate that xRAG achieves an average improvement of over 10% across six knowledge-intensive tasks, adaptable to various language model backbones, ranging from a dense 7B model to an 8x7B Mixture of Experts configuration. xRAG not only significantly outperforms previous context compression methods but also matches the performance of uncompressed models on several datasets, while reducing overall FLOPs by a factor of 3. 53. Our work pioneers new directions in retrieval-augmented generation from the perspective of multimodality fusion, and we hope it lays the foundation for future efficient and scalable retrieval-augmented systems.

PDF Details DOI

YNICL Journal 2023 Journal Article

Effect of corticosubcortical iron deposition on dysfunction in CADASIL is mediated by white matter microstructural damage

Xiuqin Jia
Yingying Li
Yunqing Ying
Xuejia Jia
Weijun Tang
Yueyan Bian
Jiajia Zhang
Danny J.J. Wang

Iron dysregulation may attenuate cognitive performance in patients with CADASIL. However, the underlying pathophysiological mechanisms remain incompletely understood. Whether white matter microstructural changes mediate these processes is largely unclear. In the present study, 30 cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) patients were confirmed via genetic analysis and 30 sex- and age-matched healthy controls underwent multimodal MRI examinations and neuropsychological assessments. Quantitative susceptibility mapping and peak width of skeletonized mean diffusivity (PSMD) were analyzed. Mediation effect analysis was performed to explore the interrelationship between iron deposition, white matter microstructural changes and cognitive deficits in CADASIL. Cognitive deterioration was most affected in memory and executive function, followed by attention and working memory in CADASIL. Excessive iron in the temporal-precuneus pathway and deep gray matter specific to CADASIL were identified. Mediation analysis further revealed that PSMD mediated the relationship between iron concentration and cognitive profile in CADASIL. The present findings provide a new perspective on iron deposition in the corticosubcortical circuit and its contribution to disease-related selective cognitive decline, in which iron concentration may affect cognition by white matter microstructural changes in CADASIL.

Details DOI

NeurIPS Conference 2023 Conference Paper

Lift Yourself Up: Retrieval-augmented Text Generation with Self-Memory

Xin Cheng
Di Luo
Xiuying Chen
Lemao Liu
Dongyan Zhao
Rui Yan

With direct access to human-written reference as memory, retrieval-augmented generation has achieved much progress in a wide range of text generation tasks. Since better memory would typically prompt better generation (we define this as primal problem). The traditional approach for memory retrieval involves selecting memory that exhibits the highest similarity to the input. However, this method is constrained by the quality of the fixed corpus from which memory is retrieved. In this paper, by exploring the duality of the primal problem: better generation also prompts better memory, we propose a novel framework, selfmem, which addresses this limitation by iteratively employing a retrieval-augmented generator to create an unbounded memory pool and using a memory selector to choose one output as memory for the subsequent generation round. This enables the model to leverage its own output, referred to as self-memory, for improved generation. We evaluate the effectiveness of selfmem on three distinct text generation tasks: neural machine translation, abstractive text summarization, and dialogue generation, under two generation paradigms: fine-tuned small model and few-shot LLM. Our approach achieves state-of-the-art results in four directions in JRC-Acquis translation dataset, 50. 3 ROUGE-1 in XSum, and 62. 9 ROUGE-1 in BigPatent, demonstrating the potential of self-memory in enhancing retrieval-augmented generation models. Furthermore, we conduct thorough analyses of each component in the selfmem framework to identify current system bottlenecks and provide insights for future research.

PDF Details

AAAI Conference 2023 Conference Paper

Partial-Label Regression

Xin Cheng
Deng-Bao Wang
Lei Feng
Min-Ling Zhang
Bo An

Partial-label learning is a popular weakly supervised learning setting that allows each training example to be annotated with a set of candidate labels. Previous studies on partial-label learning only focused on the classification setting where candidate labels are all discrete, which cannot handle continuous labels with real values. In this paper, we provide the first attempt to investigate partial-label regression, where each training example is annotated with a set of real-valued candidate labels. To solve this problem, we first propose a simple baseline method that takes the average loss incurred by candidate labels as the predictive loss. The drawback of this method lies in that the loss incurred by the true label may be overwhelmed by other false labels. To overcome this drawback, we propose an identification method that takes the least loss incurred by candidate labels as the predictive loss. We further improve it by proposing a progressive identification method to differentiate candidate labels using progressively updated weights for incurred losses. We prove that the latter two methods are model-consistent and provide convergence analysis showing the optimal parametric convergence rate. Our proposed methods are theoretically grounded and can be compatible with any models, optimizers, and losses. Experiments validate the effectiveness of our proposed methods.

PDF Details DOI

NeurIPS Conference 2023 Conference Paper

Regression with Cost-based Rejection

Xin Cheng
Yuzhou Cao
Haobo Wang
Hongxin Wei
Bo An
Lei Feng

Learning with rejection is an important framework that can refrain from making predictions to avoid critical mispredictions by balancing between prediction and rejection. Previous studies on cost-based rejection only focused on the classification setting, which cannot handle the continuous and infinite target space in the regression setting. In this paper, we investigate a novel regression problem called regression with cost-based rejection, where the model can reject to make predictions on some examples given certain rejection costs. To solve this problem, we first formulate the expected risk for this problem and then derive the Bayes optimal solution, which shows that the optimal model should reject to make predictions on the examples whose variance is larger than the rejection cost when the mean squared error is used as the evaluation metric. Furthermore, we propose to train the model by a surrogate loss function that considers rejection as binary classification and we provide conditions for the model consistency, which implies that the Bayes optimal solution can be recovered by our proposed surrogate loss. Extensive experiments demonstrate the effectiveness of our proposed method.

PDF Details

IJCAI Conference 2022 Conference Paper

Graph-based Dynamic Word Embeddings

Yuyin Lu
Xin Cheng
Ziran Liang
Yanghui Rao

As time goes by, language evolves with word semantics changing. Unfortunately, traditional word embedding methods neglect the evolution of language and assume that word representations are static. Although contextualized word embedding models can capture the diverse representations of polysemous words, they ignore temporal information as well. To tackle the aforementioned challenges, we propose a graph-based dynamic word embedding (GDWE) model, which focuses on capturing the semantic drift of words continually. We introduce word-level knowledge graphs (WKGs) to store short-term and long-term knowledge. WKGs can provide rich structural information as supplement of lexical information, which help enhance the word embedding quality and capture semantic drift quickly. Theoretical analysis and extensive experiments validate the effectiveness of our GDWE on dynamic word embedding learning.

PDF Details DOI

YNICL Journal 2019 Journal Article

18F-FDG-PET-based radiomics features to distinguish primary central nervous system lymphoma from glioblastoma

Ziren Kong
Chendan Jiang
Ruizhe Zhu
Shi Feng
Yaning Wang
Jiatong Li
Wenlin Chen
Penghao Liu

The differential diagnosis of primary central nervous system lymphoma from glioblastoma multiforme (GBM) is essential due to the difference in treatment strategies. This study retrospectively reviewed 77 patients (24 with lymphoma and 53 with GBM) to identify the stable and distinguishable characteristics of lymphoma and GBM in 18F-fluorodeocxyglucose (FDG) positron emission tomography (PET) images using a radiomics approach. Three groups of maps, namely, a standardized uptake value (SUV) map, an SUV map calibrated with the normal contralateral cortex (ncc) activity (SUV/ncc map), and an SUV map calibrated with the normal brain mean (nbm) activity (SUV/nbm map), were generated, and a total of 107 radiomics features were extracted from each SUV map. The margins of the ROI were adjusted to assess the stability of the features, and the area under the curve (AUC) of the receiver operating characteristic curve of each feature was compared with the SUVmax to evaluate the distinguishability of the features. Nighty-five radiomics features from the SUV map were significantly different between lymphoma and GBM, 46 features were numeric stable after marginal adjustment, and 31 features displayed better performance than SUVmax. Features extracted from the SUV map demonstrated higher AUCs than features from the further calibrated maps. Tumors with solid metabolic patterns were also separately evaluated and revealed similar results. Thirteen radiomics features that were stable and distinguishable than SUVmax in every circumstance were selected to distinguish lymphoma from glioblastoma, and they suggested that lymphoma has a higher SUV in most interval segments and is more mathematically heterogeneous than GBM. This study suggested that 18F-FDG-PET-based radiomics is a reliable noninvasive method to distinguish lymphoma and GBM.

Details DOI

YNICL Journal 2018 Journal Article

Voxel-based comparison of brain glucose metabolism between patients with Cushing's disease and healthy subjects

Shuai Liu
Yinyan Wang
Kaibin Xu
Fan Ping
Fang Li
Renzhi Wang
Xin Cheng

F]-fluorodeoxyglucose positron emission tomography (FDG PET), between 92 patients with CD and 118 normal subjects on a voxel-wise basis. Pearson correlation was performed to evaluate the association between cerebral FDG uptake and serum cortisol level in patients with CD. We demonstrated that certain brain regions in patients with CD showed significantly increased FDG uptake, including the basal ganglia, anteromedial temporal lobe, thalamus, precentral cortex, and cerebellum. The clusters that demonstrated significantly decreased uptake were mainly located in the medial and lateral frontal cortex, superior and inferior parietal lobule, medial occipital cortex, and insular cortex. The metabolic rate of the majority of these regions was found to be significantly correlated with the serum cortisol level. Our findings may help to explain the underlying mechanisms of cognitive impairment and psychiatric symptoms in patients exposed to excessive glucocorticoids and evaluate the efficacy of treatments during follow-up.

Details DOI

YNICL Journal 2016 Journal Article

Brain glucose metabolism is associated with hormone level in Cushing's disease: A voxel-based study using FDG-PET

Shuai Liu
Yinyan Wang
Kaibin Xu
Fan Ping
Renzhi Wang
Fang Li
Xin Cheng

Chronic exposure to elevated levels of glucocorticoids can exert a neurotoxic effect in patients, possibly manifesting as molecular imaging alterations in patients. The aim of this study was to investigate the potential association between brain metabolism and elevated hormone level using (18)F-fluorodeoxyglucose positron emission tomography. We retrospectively enrolled 92 consecutive patients with confirmed diagnosis of Cushing's disease. A voxel-based analysis was performed to investigate the association between cerebral (18)F-fluorodeoxyglucose uptake and serum cortisol level. Relatively impaired metabolism of specific brain regions correlated with serum cortisol level was found. Specifically, notable correlations were found in the hippocampus, amygdala, and cerebellum, regions considered to be involved in the regulation and central action of glucocorticoids. Moreover, some hormone-associated regions were found in the frontal and occipital cortex, possibly mediating the cognitive changes seen in Cushing's disease. Our findings link patterns of perturbed brain metabolism relates to individual hormone level, thus presenting a substrate for cognitive disturbances seen in Cushing's disease patients, as well as in other conditions with abnormal cortisol levels.

Details DOI