Phi Le Nguyen

Thanh Tam Nguyen, Thanh Trung Huynh, Zhao Ren et al.

Today, computer systems hold large amounts of personal data. Yet while such an abundance of data allows breakthroughs in artificial intelligence, and especially machine learning (ML), its existence can be a threat to user privacy, and it can weaken the bonds of trust between humans and AI. Recent regulations now require that, on request, private information about a user must be removed from both computer systems and from ML models, i.e. ``the right to be forgotten''). While removing data from back-end databases should be straightforward, it is not sufficient in the AI context as ML models often `remember' the old data. Contemporary adversarial attacks on trained models have proven that we can learn whether an instance or an attribute belonged to the training data. This phenomenon calls for a new paradigm, namely machine unlearning, to make ML models forget about particular data. It turns out that recent works on machine unlearning have not been able to completely solve the problem due to the lack of common frameworks and resources. Therefore, this paper aspires to present a comprehensive examination of machine unlearning's concepts, scenarios, methods, and applications. Specifically, as a category collection of cutting-edge studies, the intention behind this article is to serve as a comprehensive resource for researchers and practitioners seeking an introduction to machine unlearning and its formulations, design criteria, removal requests, algorithms, and applications. In addition, we aim to highlight the key findings, current trends, and new research areas that have not yet featured the use of machine unlearning but could benefit greatly from it. We hope this survey serves as a valuable resource for ML researchers and those seeking to innovate privacy technologies. Our resources are publicly available at https://github.com/tamlhp/awesome-machine-unlearning.

Duy A. Nguyen, Trang H. Tran, Huy Hieu Pham et al. · cmu, deepmind

In this work, we investigate the time series representation learning problem using self-supervised techniques. Contrastive learning is well-known in this area as it is a powerful method for extracting information from the series and generating task-appropriate representations. Despite its proficiency in capturing time series characteristics, these techniques often overlook a critical factor - the inherent noise in this type of data, a consideration usually emphasized in general time series analysis. Moreover, there is a notable absence of attention to developing efficient yet lightweight encoder architectures, with an undue focus on delivering contrastive losses. Our work address these gaps by proposing an innovative training strategy that promotes consistent representation learning, accounting for the presence of noise-prone signals in natural time series. Furthermore, we propose an encoder architecture that incorporates dilated convolution within the Inception block, resulting in a scalable and robust network with a wide receptive field. Experimental findings underscore the effectiveness of our method, consistently outperforming state-of-the-art approaches across various tasks, including forecasting, classification, and abnormality detection. Notably, our method attains the top rank in over two-thirds of the classification UCR datasets, utilizing only 40% of the parameters compared to the second-best approach. Our source code for CoInception framework is accessible at https://github.com/anhduy0911/CoInception.

Thanh Trung Huynh, Minh Hieu Nguyen, Thanh Tam Nguyen et al.

Advances in deep neural network (DNN) architectures have enabled new prediction techniques for stock market data. Unlike other multivariate time-series data, stock markets show two unique characteristics: (i) \emph{multi-order dynamics}, as stock prices are affected by strong non-pairwise correlations (e.g., within the same industry); and (ii) \emph{internal dynamics}, as each individual stock shows some particular behaviour. Recent DNN-based methods capture multi-order dynamics using hypergraphs, but rely on the Fourier basis in the convolution, which is both inefficient and ineffective. In addition, they largely ignore internal dynamics by adopting the same model for each stock, which implies a severe information loss. In this paper, we propose a framework for stock movement prediction to overcome the above issues. Specifically, the framework includes temporal generative filters that implement a memory-based mechanism onto an LSTM network in an attempt to learn individual patterns per stock. Moreover, we employ hypergraph attentions to capture the non-pairwise correlations. Here, using the wavelet basis instead of the Fourier basis, enables us to simplify the message passing and focus on the localized convolution. Experiments with US market data over six years show that our framework outperforms state-of-the-art methods in terms of profit and stability. Our source code and data are available at \url{https://github.com/thanhtrunghuynh93/estimate}.

Quan Nguyen, Hieu H. Pham, Kok-Seng Wong et al. · cmu, deepmind

We introduce FedDCT, a novel distributed learning paradigm that enables the usage of large, high-performance CNNs on resource-limited edge devices. As opposed to traditional FL approaches, which require each client to train the full-size neural network independently during each training round, the proposed FedDCT allows a cluster of several clients to collaboratively train a large deep learning model by dividing it into an ensemble of several small sub-models and train them on multiple devices in parallel while maintaining privacy. In this collaborative training process, clients from the same cluster can also learn from each other, further improving their ensemble performance. In the aggregation stage, the server takes a weighted average of all the ensemble models trained by all the clusters. FedDCT reduces the memory requirements and allows low-end devices to participate in FL. We empirically conduct extensive experiments on standardized datasets, including CIFAR-10, CIFAR-100, and two real-world medical datasets HAM10000 and VAIPE. Experimental results show that FedDCT outperforms a set of current SOTA FL methods with interesting convergence behaviors. Furthermore, compared to other existing approaches, FedDCT achieves higher accuracy and substantially reduces the number of communication rounds (with $4-8$ times fewer memory requirements) to achieve the desired accuracy on the testing dataset without incurring any extra training cost on the server side.

Darnbi Sakong, Viet Hung Vu, Thanh Trung Huynh et al.

Recent advancements in recommender systems have focused on integrating knowledge graphs (KGs) to leverage their auxiliary information. The core idea of KG-enhanced recommenders is to incorporate rich semantic information for more accurate recommendations. However, two main challenges persist: i) Neglecting complex higher-order interactions in the KG-based user-item network, potentially leading to sub-optimal recommendations, and ii) Dealing with the heterogeneous modalities of input sources, such as user-item bipartite graphs and KGs, which may introduce noise and inaccuracies. To address these issues, we present a novel Knowledge-enhanced Heterogeneous Hypergraph Recommender System (KHGRec). KHGRec captures group-wise characteristics of both the interaction network and the KG, modeling complex connections in the KG. Using a collaborative knowledge heterogeneous hypergraph (CKHG), it employs two hypergraph encoders to model group-wise interdependencies and ensure explainability. Additionally, it fuses signals from the input graphs with cross-view self-supervised learning and attention mechanisms. Extensive experiments on four real-world datasets show our model's superiority over various state-of-the-art baselines, with an average 5.18\% relative improvement. Additional tests on noise resilience, missing data, and cold-start problems demonstrate the robustness of our KHGRec framework. Our model and evaluation datasets are publicly available at \url{https://github.com/viethungvu1998/KHGRec}.

Thuy Dung Nguyen, Tuan Nguyen, Phi Le Nguyen et al. · baidu

Federated learning (FL) is a machine learning (ML) approach that allows the use of distributed data without compromising personal privacy. However, the heterogeneous distribution of data among clients in FL can make it difficult for the orchestration server to validate the integrity of local model updates, making FL vulnerable to various threats, including backdoor attacks. Backdoor attacks involve the insertion of malicious functionality into a targeted model through poisoned updates from malicious clients. These attacks can cause the global model to misbehave on specific inputs while appearing normal in other cases. Backdoor attacks have received significant attention in the literature due to their potential to impact real-world deep learning applications. However, they have not been thoroughly studied in the context of FL. In this survey, we provide a comprehensive survey of current backdoor attack strategies and defenses in FL, including a comprehensive analysis of different approaches. We also discuss the challenges and potential future directions for attacks and defenses in the context of FL.

Trong-Tung Nguyen, Hieu H. Pham, Phi Le Nguyen et al.

Classifying pill categories from real-world images is crucial for various smart healthcare applications. Although existing approaches in image classification might achieve a good performance on fixed pill categories, they fail to handle novel instances of pill categories that are frequently presented to the learning algorithm. To this end, a trivial solution is to train the model with novel classes. However, this may result in a phenomenon known as catastrophic forgetting, in which the system forgets what it learned in previous classes. In this paper, we address this challenge by introducing the class incremental learning (CIL) ability to traditional pill image classification systems. Specifically, we propose a novel incremental multi-stream intermediate fusion framework enabling incorporation of an additional guidance information stream that best matches the domain of the problem into various state-of-the-art CIL methods. From this framework, we consider color-specific information of pill images as a guidance stream and devise an approach, namely "Color Guidance with Multi-stream intermediate fusion"(CG-IMIF) for solving CIL pill image classification task. We conduct comprehensive experiments on real-world incremental pill image classification dataset, namely VAIPE-PCIL, and find that the CG-IMIF consistently outperforms several state-of-the-art methods by a large margin in different task settings. Our code, data, and trained model are available at https://github.com/vinuni-vishc/CG-IMIF.

Nang Hung Nguyen, Phi Le Nguyen, Duc Long Nguyen et al.

The uneven distribution of local data across different edge devices (clients) results in slow model training and accuracy reduction in federated learning. Naive federated learning (FL) strategy and most alternative solutions attempted to achieve more fairness by weighted aggregating deep learning models across clients. This work introduces a novel non-IID type encountered in real-world datasets, namely cluster-skew, in which groups of clients have local data with similar distributions, causing the global model to converge to an over-fitted solution. To deal with non-IID data, particularly the cluster-skewed data, we propose FedDRL, a novel FL model that employs deep reinforcement learning to adaptively determine each client's impact factor (which will be used as the weights in the aggregation process). Extensive experiments on a suite of federated datasets confirm that the proposed FedDRL improves favorably against FedAvg and FedProx methods, e.g., up to 4.05% and 2.17% on average for the CIFAR-100 dataset, respectively.

Cong Huy Nguyen, Son Dinh Nguyen, Guanlin Li et al.

Automated medical report generation for 3D PET/CT imaging is fundamentally challenged by the high-dimensional nature of volumetric data and a critical scarcity of annotated datasets, particularly for low-resource languages. Current black-box methods map whole volumes to reports, ignoring the clinical workflow of analyzing localized Regions of Interest (RoIs) to derive diagnostic conclusions. In this paper, we bridge this gap by introducing VietPET-RoI, the first large-scale 3D PET/CT dataset with fine-grained RoI annotation for a low-resource language, comprising 600 PET/CT samples and 1,960 manually annotated RoIs, paired with corresponding clinical reports. Furthermore, to demonstrate the utility of this dataset, we propose HiRRA, a novel framework that mimics the professional radiologist diagnostic workflow by employing graph-based relational modules to capture dependencies between RoI attributes. This approach shifts from global pattern matching toward localized clinical findings. Additionally, we introduce new clinical evaluation metrics, namely RoI Coverage and RoI Quality Index, that measure both RoI localization accuracy and attribute description fidelity using LLM-based extraction. Extensive evaluation demonstrates that our framework achieves SOTA performance, surpassing existing models by 19.7% in BLEU and 4.7% in ROUGE-L, while achieving a remarkable 45.8% improvement in clinical metrics, indicating enhanced clinical reliability and reduced hallucination. Our code and dataset are available on GitHub.

Trung Thanh Nguyen, Hoang Dang Nguyen, Thanh Hung Nguyen et al.

Medication mistaking is one of the risks that can result in unpredictable consequences for patients. To mitigate this risk, we develop an automatic system that correctly identifies pill-prescription from mobile images. Specifically, we define a so-called pill-prescription matching task, which attempts to match the images of the pills taken with the pills' names in the prescription. We then propose PIMA, a novel approach using Graph Neural Network (GNN) and contrastive learning to address the targeted problem. In particular, GNN is used to learn the spatial correlation between the text boxes in the prescription and thereby highlight the text boxes carrying the pill names. In addition, contrastive learning is employed to facilitate the modeling of cross-modal similarity between textual representations of pill names and visual representations of pill images. We conducted extensive experiments and demonstrated that PIMA outperforms baseline models on a real-world dataset of pill and prescription images that we constructed. Specifically, PIMA improves the accuracy from 19.09% to 46.95% compared to other baselines. We believe our work can open up new opportunities to build new clinical applications and improve medication safety and patient care.

Anh Duy Nguyen, Huy Hieu Pham, Huynh Thanh Trung et al.

Due to the significant resemblance in visual appearance, pill misuse is prevalent and has become a critical issue, responsible for one-third of all deaths worldwide. Pill identification, thus, is a crucial concern needed to be investigated thoroughly. Recently, several attempts have been made to exploit deep learning to tackle the pill identification problem. However, most published works consider only single-pill identification and fail to distinguish hard samples with identical appearances. Also, most existing pill image datasets only feature single pill images captured in carefully controlled environments under ideal lighting conditions and clean backgrounds. In this work, we are the first to tackle the multi-pill detection problem in real-world settings, aiming at localizing and identifying pills captured by users in a pill intake. Moreover, we also introduce a multi-pill image dataset taken in unconstrained conditions. To handle hard samples, we propose a novel method for constructing heterogeneous a priori graphs incorporating three forms of inter-pill relationships, including co-occurrence likelihood, relative size, and visual semantic correlation. We then offer a framework for integrating a priori with pills' visual features to enhance detection accuracy. Our experimental results have proved the robustness, reliability, and explainability of the proposed framework. Experimentally, it outperforms all detection benchmarks in terms of all evaluation metrics. Specifically, our proposed framework improves COCO mAP metrics by 9.4% over Faster R-CNN and 12.0% compared to vanilla YOLOv5. Our study opens up new opportunities for protecting patients from medication errors using an AI-based pill identification solution.

Nang Hung Nguyen, Duc Long Nguyen, Trong Bang Nguyen et al.

Federated learning enables edge devices to train a global model collaboratively without exposing their data. Despite achieving outstanding advantages in computing efficiency and privacy protection, federated learning faces a significant challenge when dealing with non-IID data, i.e., data generated by clients that are typically not independent and identically distributed. In this paper, we tackle a new type of Non-IID data, called cluster-skewed non-IID, discovered in actual data sets. The cluster-skewed non-IID is a phenomenon in which clients can be grouped into clusters with similar data distributions. By performing an in-depth analysis of the behavior of a classification model's penultimate layer, we introduce a metric that quantifies the similarity between two clients' data distributions without violating their privacy. We then propose an aggregation scheme that guarantees equality between clusters. In addition, we offer a novel local training regularization based on the knowledge-distillation technique that reduces the overfitting problem at clients and dramatically boosts the training scheme's performance. We theoretically prove the superiority of the proposed aggregation over the benchmark FedAvg. Extensive experimental results on both standard public datasets and our in-house real-world dataset demonstrate that the proposed approach improves accuracy by up to 16% compared to the FedAvg algorithm.

Ngoc Bui, Phi Le Nguyen, Viet Anh Nguyen et al.

Wireless sensor networks consist of randomly distributed sensor nodes for monitoring targets or areas of interest. Maintaining the network for continuous surveillance is a challenge due to the limited battery capacity in each sensor. Wireless power transfer technology is emerging as a reliable solution for energizing the sensors by deploying a mobile charger (MC) to recharge the sensor. However, designing an optimal charging path for the MC is challenging because of uncertainties arising in the networks. The energy consumption rate of the sensors may fluctuate significantly due to unpredictable changes in the network topology, such as node failures. These changes also lead to shifts in the importance of each sensor, which are often assumed to be the same in existing works. We address these challenges in this paper by proposing a novel adaptive charging scheme using a deep reinforcement learning (DRL) approach. Specifically, we endow the MC with a charging policy that determines the next sensor to charge conditioning on the current state of the network. We then use a deep neural network to parametrize this charging policy, which will be trained by reinforcement learning techniques. Our model can adapt to spontaneous changes in the network topology. The empirical results show that the proposed algorithm outperforms the existing on-demand algorithms by a significant margin.

Thuy Dung Nguyen, Anh Duy Nguyen, Kok-Seng Wong et al.

Federated learning (FL) enables multiple clients to train a model without compromising sensitive data. The decentralized nature of FL makes it susceptible to adversarial attacks, especially backdoor insertion during training. Recently, the edge-case backdoor attack employing the tail of the data distribution has been proposed as a powerful one, raising questions about the shortfall in current defenses' robustness guarantees. Specifically, most existing defenses cannot eliminate edge-case backdoor attacks or suffer from a trade-off between backdoor-defending effectiveness and overall performance on the primary task. To tackle this challenge, we propose FedGrad, a novel backdoor-resistant defense for FL that is resistant to cutting-edge backdoor attacks, including the edge-case attack, and performs effectively under heterogeneous client data and a large number of compromised clients. FedGrad is designed as a two-layer filtering mechanism that thoroughly analyzes the ultimate layer's gradient to identify suspicious local updates and remove them from the aggregation process. We evaluate FedGrad under different attack scenarios and show that it significantly outperforms state-of-the-art defense mechanisms. Notably, FedGrad can almost 100% correctly detect the malicious participants, thus providing a significant reduction in the backdoor effect (e.g., backdoor accuracy is less than 8%) while not reducing the main accuracy on the primary task.

Anh Duy Nguyen, Thuy Dung Nguyen, Huy Hieu Pham et al.

Identifying pills given their captured images under various conditions and backgrounds has been becoming more and more essential. Several efforts have been devoted to utilizing the deep learning-based approach to tackle the pill recognition problem in the literature. However, due to the high similarity between pills' appearance, misrecognition often occurs, leaving pill recognition a challenge. To this end, in this paper, we introduce a novel approach named PIKA that leverages external knowledge to enhance pill recognition accuracy. Specifically, we address a practical scenario (which we call contextual pill recognition), aiming to identify pills in a picture of a patient's pill intake. Firstly, we propose a novel method for modeling the implicit association between pills in the presence of an external data source, in this case, prescriptions. Secondly, we present a walk-based graph embedding model that transforms from the graph space to vector space and extracts condensed relational features of the pills. Thirdly, a final framework is provided that leverages both image-based visual and graph-based relational features to accomplish the pill identification task. Within this framework, the visual representation of each pill is mapped to the graph embedding space, which is then used to execute attention over the graph representation, resulting in a semantically-rich context vector that aids in the final classification. To our knowledge, this is the first study to use external prescription data to establish associations between medicines and to classify them using this aiding information. The architecture of PIKA is lightweight and has the flexibility to incorporate into any recognition backbones. The experimental results show that by leveraging the external knowledge graph, PIKA can improve the recognition accuracy from 4.8% to 34.1% in terms of F1-score, compared to baselines.

Nam Nguyen Phuong, Duc Nguyen The Minh, Phi Le Nguyen et al.

Conventional test-time adaptation (TTA) approaches typically adapt the model using only a small fraction of test samples, often those with low-entropy predictions, thereby failing to fully leverage the available information in the test distribution. This paper introduces DualTTA, a novel framework that improves performance under distribution shifts by utilizing a larger and more diverse set of test samples. DualTTA identifies two distinct groups: one where the model's predictions are likely consistent with the underlying semantics, and another where predictions are likely incorrect. For the first group, it minimizes prediction entropy to reinforce reliable decisions; for the second, it maximizes entropy to suppress overconfident errors and unlearn spurious behavior. These groups are adaptively selected using a new reliability criterion that measures prediction stability under both semantic-preserving and semantic-altering transformations, addressing the limitations of purely entropy-based selection. We further provide theoretical analysis and empirical justification showing that our approach enables a tighter separation between reliable and unreliable samples, in the context of their suitability for adaptation, leading to provably more effective model updates.

Tue M. Cao, Nhat H. Tran, Phi Le Nguyen et al.

This work discusses the use of contrastive learning and deep learning for diagnosing cardiovascular diseases from electrocardiography (ECG) signals. While the ECG signals usually contain 12 leads (channels), many healthcare facilities and devices lack access to all these 12 leads. This raises the problem of how to use only fewer ECG leads to produce meaningful diagnoses with high performance. We introduce a simple experiment to test whether contrastive learning can be applied to this task. More specifically, we added the similarity between the embedding vectors when the 12 leads signal and the fewer leads ECG signal to the loss function to bring these representations closer together. Despite its simplicity, this has been shown to have improved the performance of diagnosing with all lead combinations, proving the potential of contrastive learning on this task.

Thanh Tam Nguyen, Thanh Trung Huynh, Zhao Ren et al.

As the adoption of explainable AI (XAI) continues to expand, the urgency to address its privacy implications intensifies. Despite a growing corpus of research in AI privacy and explainability, there is little attention on privacy-preserving model explanations. This article presents the first thorough survey about privacy attacks on model explanations and their countermeasures. Our contribution to this field comprises a thorough analysis of research papers with a connected taxonomy that facilitates the categorisation of privacy attacks and countermeasures based on the targeted explanations. This work also includes an initial investigation into the causes of privacy leaks. Finally, we discuss unresolved issues and prospective research directions uncovered in our analysis. This survey aims to be a valuable resource for the research community and offers clear insights for those new to this domain. To support ongoing research, we have established an online resource repository, which will be continuously updated with new and relevant findings. Interested readers are encouraged to access our repository at https://github.com/tamlhp/awesome-privex.

Thu Hang Phung, Duong M. Nguyen, Thanh Trung Huynh et al.

This paper introduces a generalized federated prompt-tuning framework for practical scenarios where local datasets are multi-modal and exhibit different distributional patterns of missing features at the input level. The proposed framework bridges the gap between federated learning and multi-modal prompt-tuning which have traditionally focused on either uni-modal or centralized data. A key challenge in this setting arises from the lack of semantic alignment between prompt instructions that encode similar distributional patterns of missing data across different clients. To address this, our framework introduces specialized client-tuning and server-aggregation designs that simultaneously optimize, align, and aggregate prompt-tuning instructions across clients and data modalities. This allows prompt instructions to complement one another and be combined effectively. Extensive evaluations on diverse multimodal benchmark datasets demonstrate that our work consistently outperforms state-of-the-art (SOTA) baselines.

Manh Cuong Dao, Phi Le Nguyen, Thao Nguyen Truong et al.

Offline optimization has recently emerged as an increasingly popular approach to mitigate the prohibitively expensive cost of online experimentation. The key idea is to learn a surrogate of the black-box function that underlines the target experiment using a static (offline) dataset of its previous input-output queries. Such an approach is, however, fraught with an out-of-distribution issue where the learned surrogate becomes inaccurate outside the offline data regimes. To mitigate this, existing offline optimizers have proposed numerous conditioning techniques to prevent the learned surrogate from being too erratic. Nonetheless, such conditioning strategies are often specific to particular surrogate or search models, which might not generalize to a different model choice. This motivates us to develop a model-agnostic approach instead, which incorporates a notion of model sharpness into the training loss of the surrogate as a regularizer. Our approach is supported by a new theoretical analysis demonstrating that reducing surrogate sharpness on the offline dataset provably reduces its generalized sharpness on unseen data. Our analysis extends existing theories from bounding generalized prediction loss (on unseen data) with loss sharpness to bounding the worst-case generalized surrogate sharpness with its empirical estimate on training data, providing a new perspective on sharpness regularization. Our extensive experimentation on a diverse range of optimization tasks also shows that reducing surrogate sharpness often leads to significant improvement, marking (up to) a noticeable 9.6% performance boost. Our code is publicly available at https://github.com/cuong-dm/IGNITE

Thanh Tam Nguyen, Zhao Ren, Trinh Pham et al.

The rapid advancement of large language models (LLMs) and multimodal learning has transformed digital content creation and manipulation. Traditional visual editing tools require significant expertise, limiting accessibility. Recent strides in instruction-based editing have enabled intuitive interaction with visual content, using natural language as a bridge between user intent and complex editing operations. This survey provides an overview of these techniques, focusing on how LLMs and multimodal models empower users to achieve precise visual modifications without deep technical knowledge. By synthesizing over 100 publications, we explore methods from generative adversarial networks to diffusion models, examining multimodal integration for fine-grained content control. We discuss practical applications across domains such as fashion, 3D scene manipulation, and video synthesis, highlighting increased accessibility and alignment with human intuition. Our survey compares existing literature, emphasizing LLM-empowered editing, and identifies key challenges to stimulate further research. We aim to democratize powerful visual editing across various industries, from entertainment to education. Interested readers are encouraged to access our repository at https://github.com/tamlhp/awesome-instruction-editing.

Dac Thai Nguyen, Trung Thanh Nguyen, Huu Tien Nguyen et al.

Positron Emission Tomography (PET) and Computed Tomography (CT) are essential for diagnosing, staging, and monitoring various diseases, particularly cancer. Despite their importance, the use of PET/CT systems is limited by the necessity for radioactive materials, the scarcity of PET scanners, and the high cost associated with PET imaging. In contrast, CT scanners are more widely available and significantly less expensive. In response to these challenges, our study addresses the issue of generating PET images from CT images, aiming to reduce both the medical examination cost and the associated health risks for patients. Our contributions are twofold: First, we introduce a conditional diffusion model named CPDM, which, to our knowledge, is one of the initial attempts to employ a diffusion model for translating from CT to PET images. Second, we provide the largest CT-PET dataset to date, comprising 2,028,628 paired CT-PET images, which facilitates the training and evaluation of CT-to-PET translation models. For the CPDM model, we incorporate domain knowledge to develop two conditional maps: the Attention map and the Attenuation map. The former helps the diffusion process focus on areas of interest, while the latter improves PET data correction and ensures accurate diagnostic information. Experimental evaluations across various benchmarks demonstrate that CPDM surpasses existing methods in generating high-quality PET images in terms of multiple metrics. The source code and data samples are available at https://github.com/thanhhff/CPDM.

Huu Tien Nguyen, Dac Thai Nguyen, The Minh Duc Nguyen et al.

Vision-Language Foundation Models (VLMs), trained on large-scale multimodal datasets, have driven significant advances in Artificial Intelligence (AI) by enabling rich cross-modal reasoning. Despite their success in general domains, applying these models to medical imaging remains challenging due to the limited availability of diverse imaging modalities and multilingual clinical data. Most existing medical VLMs are trained on a subset of imaging modalities and focus primarily on high-resource languages, thus limiting their generalizability and clinical utility. To address these limitations, we introduce a novel Vietnamese-language multimodal medical dataset consisting of 2,757 whole-body PET/CT volumes from independent patients and their corresponding full-length clinical reports. This dataset is designed to fill two pressing gaps in medical AI development: (1) the lack of PET/CT imaging data in existing VLMs training corpora, which hinders the development of models capable of handling functional imaging tasks; and (2) the underrepresentation of low-resource languages, particularly the Vietnamese language, in medical vision-language research. To the best of our knowledge, this is the first dataset to provide comprehensive PET/CT-report pairs in Vietnamese. We further introduce a training framework to enhance VLMs' learning, including data augmentation and expert-validated test sets. We conduct comprehensive experiments benchmarking state-of-the-art VLMs on downstream tasks. The experimental results show that incorporating our dataset significantly improves the performance of existing VLMs. We believe this dataset and benchmark will serve as a pivotal step in advancing the development of more robust VLMs for medical imaging, especially for low-resource languages and clinical use in Vietnamese healthcare. The source code is available at https://github.com/AIoT-Lab-BKAI/ViPET-ReportGen.

Minh Khoi Nguyen, Dai Lam Le, Amir Reza Jafari et al.

Large vision-language models (VLMs) demonstrate strong performance in medical image understanding, but frequently generate clinically plausible yet incorrect statements, raising significant safety concerns. Existing medical hallucination benchmarks primarily focus on 2D imaging with one-shot diagnostic questions, offering limited insight into whether predictions are grounded in correct localization and abnormality identification, allowing critical reasoning errors to remain hidden behind seemingly correct diagnoses. We introduce Med-StepBench, the first large-scale benchmark for step-wise hallucination detection in 3D oncological PET/CT, comprising over 12,000 images and more than 1,000,000 image-statement pairs across volumetric and multi-view 2D data, which decomposes clinical reasoning into four expert-designed diagnostic stages. Using clinician-verified annotations, we perform the first step-level evaluation of general-purpose and medical VLMs, revealing systematic failure modes obscured by aggregate accuracy metrics. Furthermore, we show that current VLMs are highly susceptible to adversarial yet clinically plausible intermediate explanations, which significantly amplify hallucinations despite contradictory visual evidence. Together, our findings highlight fundamental limitations in grounding multi-step clinical reasoning and establish Med-StepBench as a rigorous benchmark for developing safer and more reliable medical VLMs.

Trong Khiem Tran, Manh Cuong Dao, Phi Le Nguyen et al.

Adapting pre-trained models to unseen feature modalities has become increasingly important due to the growing need for cross-disciplinary knowledge integration.~A key challenge here is how to align the representation of new modalities with the most relevant parts of the pre-trained model's representation space to enable accurate knowledge transfer.~This requires combining feature alignment with target fine-tuning, but uncalibrated combinations can exacerbate misalignment between the source and target feature-label structures and reduce target generalization.~Existing work however lacks a theoretical understanding of this critical interaction between feature alignment and target fitting.~To bridge this gap, we develop a principled framework that establishes a provable generalization bound on the target error, which explains the interaction between feature alignment and target fitting through a novel concept of feature-label distortion.~This bound offers actionable insights into how this interaction should be optimized for practical algorithm design. The resulting approach achieves significantly improved performance over state-of-the-art methods across a wide range of benchmark datasets.

Manh Cuong Dao, Quang Hung Pham, Phi Le Nguyen et al.

Uncertainty calibration in pre-trained transformers is critical for their reliable deployment in risk-sensitive applications. Yet, most existing pre-trained transformers do not have a principled mechanism for uncertainty propagation through their feature transformation stack. In this work, we propose a diffusion-inspired reconfiguration of transformers in which each feature transformation block is modeled as a probabilistic mapping. Composing these probabilistic mappings reveals a probability path that mimics the structure of a diffusion process, transporting data mass from the input distribution to the pre-trained feature distribution. This probability path can then be recompiled on a diffusion process with a unified transition model to enable principled propagation of representation uncertainty throughout the pre-trained model's architecture while maintaining its original predictive performance. Empirical results across a variety of vision and language benchmarks demonstrate that our method achieves superior calibration and predictive accuracy compared to existing uncertainty-aware transformers.

Nang Hung Nguyen, Phi Le Nguyen, Thao Nguyen Truong et al.

This paper introduces a new framework for recovering causal graphs from observational data, leveraging the observation that the distribution of an effect, conditioned on its causes, remains invariant to changes in the prior distribution of those causes. This insight enables a direct test for potential causal relationships by checking the variance of their corresponding effect-cause conditional distributions across multiple downsampled subsets of the data. These subsets are selected to reflect different prior cause distributions, while preserving the effect-cause conditional relationships. Using this invariance test and exploiting an (empirical) sparsity of most causal graphs, we develop an algorithm that efficiently uncovers causal relationships with quadratic complexity in the number of observational variables, reducing the processing time by up to 25x compared to state-of-the-art methods. Our empirical experiments on a varied benchmark of large-scale datasets show superior or equivalent performance compared to existing works, while achieving enhanced scalability.

Duong M. Nguyen, Trong Nghia Hoang, Thanh Trung Huynh et al.

Multimodal federated learning in real-world settings often encounters incomplete and heterogeneous data across clients. This results in misaligned local feature representations that limit the effectiveness of model aggregation. Unlike prior work that assumes either differing modality sets without missing input features or a shared modality set with missing features across clients, we consider a more general and realistic setting where each client observes a different subset of modalities and might also have missing input features within each modality. To address the resulting misalignment in learned representations, we propose a new federated learning framework featuring locally adaptive representations based on learnable client-side embedding controls that encode each client's data-missing patterns. These embeddings serve as reconfiguration signals that align the globally aggregated representation with each client's local context, enabling more effective use of shared information. Furthermore, the embedding controls can be algorithmically aggregated across clients with similar data-missing patterns to enhance the robustness of reconfiguration signals in adapting the global representation. Empirical results on multiple federated multimodal benchmarks with diverse data-missing patterns across clients demonstrate the efficacy of the proposed method, achieving up to 36.45\% performance improvement under severe data incompleteness. The method is also supported by a theoretical analysis with an explicit performance bound that matches our empirical observations. Our source codes are provided at https://github.com/nmduonggg/PEPSY

Manh Cuong Dao, The Hung Tran, Phi Le Nguyen et al.

This paper studies the black-box optimization task which aims to find the maxima of a black-box function using a static set of its observed input-output pairs. This is often achieved via learning and optimizing a surrogate function with that offline data. Alternatively, it can also be framed as an inverse modeling task that maps a desired performance to potential input candidates that achieve it. Both approaches are constrained by the limited amount of offline data. To mitigate this limitation, we introduce a new perspective that casts offline optimization as a distributional translation task. This is formulated as learning a probabilistic bridge transforming an implicit distribution of low-value inputs (i.e., offline data) into another distribution of high-value inputs (i.e., solution candidates). Such probabilistic bridge can be learned using low- and high-value inputs sampled from synthetic functions that resemble the target function. These synthetic functions are constructed as the mean posterior of multiple Gaussian processes fitted with different parameterizations on the offline data, alleviating the data bottleneck. The proposed approach is evaluated on an extensive benchmark comprising most recent methods, demonstrating significant improvement and establishing a new state-of-the-art performance. Our code is publicly available at https://github.com/cuong-dm/ROOT.

Trung Kien Pham, Hoang Minh Vu, Anh Duc Chu et al.

Attenuation artifacts remain a significant challenge in cardiac Myocardial Perfusion Imaging (MPI) using Single-Photon Emission Computed Tomography (SPECT), often compromising diagnostic accuracy and reducing clinical interpretability. While hybrid SPECT/CT systems mitigate these artifacts through CT-derived attenuation maps, their high cost, limited accessibility, and added radiation exposure hinder widespread clinical adoption. In this study, we propose a novel CT-free solution to attenuation correction in cardiac SPECT. Specifically, we introduce Physics-aware Attenuation Correction Diffusion Model (PADM), a diffusion-based generative method that incorporates explicit physics priors via a teacher--student distillation mechanism. This approach enables attenuation artifact correction using only Non-Attenuation-Corrected (NAC) input, while still benefiting from physics-informed supervision during training. To support this work, we also introduce CardiAC, a comprehensive dataset comprising 424 patient studies with paired NAC and Attenuation-Corrected (AC) reconstructions, alongside high-resolution CT-based attenuation maps. Extensive experiments demonstrate that PADM outperforms state-of-the-art generative models, delivering superior reconstruction fidelity across both quantitative metrics and visual assessment.

Manh Cuong Dao, Phi Le Nguyen, Thao Nguyen Truong et al.

Offline optimization is an important task in numerous material engineering domains where online experimentation to collect data is too expensive and needs to be replaced by an in silico maximization of a surrogate of the black-box function. Although such a surrogate can be learned from offline data, its prediction might not be reliable outside the offline data regime, which happens when the surrogate has narrow prediction margin and is (therefore) sensitive to small perturbations of its parameterization. This raises the following questions: (1) how to regulate the sensitivity of a surrogate model; and (2) whether conditioning an offline optimizer with such less sensitive surrogate will lead to better optimization performance. To address these questions, we develop an optimizable sensitivity measurement for the surrogate model, which then inspires a sensitivity-informed regularizer that is applicable to a wide range of offline optimizers. This development is both orthogonal and synergistic to prior research on offline optimization, which is demonstrated in our extensive experiment benchmark.

Dung Nguyen, Minh Khoi Ho, Huy Ta et al.

Medical Large Multi-modal Models (LMMs) have demonstrated remarkable capabilities in medical data interpretation. However, these models frequently generate hallucinations contradicting source evidence, particularly due to inadequate localization reasoning. This work reveals a critical limitation in current medical LMMs: instead of analyzing relevant pathological regions, they often rely on linguistic patterns or attend to irrelevant image areas when responding to disease-related queries. To address this, we introduce HEAL-MedVQA (Hallucination Evaluation via Localization MedVQA), a comprehensive benchmark designed to evaluate LMMs' localization abilities and hallucination robustness. HEAL-MedVQA features (i) two innovative evaluation protocols to assess visual and textual shortcut learning, and (ii) a dataset of 67K VQA pairs, with doctor-annotated anatomical segmentation masks for pathological regions. To improve visual reasoning, we propose the Localize-before-Answer (LobA) framework, which trains LMMs to localize target regions of interest and self-prompt to emphasize segmented pathological areas, generating grounded and reliable answers. Experimental results demonstrate that our approach significantly outperforms state-of-the-art biomedical LMMs on the challenging HEAL-MedVQA benchmark, advancing robustness in medical VQA.

Ta Duc Huy, Duy Anh Huynh, Yutong Xie et al.

Visual grounding (VG) is the capability to identify the specific regions in an image associated with a particular text description. In medical imaging, VG enhances interpretability by highlighting relevant pathological features corresponding to textual descriptions, improving model transparency and trustworthiness for wider adoption of deep learning models in clinical practice. Current models struggle to associate textual descriptions with disease regions due to inefficient attention mechanisms and a lack of fine-grained token representations. In this paper, we empirically demonstrate two key observations. First, current VLMs assign high norms to background tokens, diverting the model's attention from regions of disease. Second, the global tokens used for cross-modal learning are not representative of local disease tokens. This hampers identifying correlations between the text and disease tokens. To address this, we introduce simple, yet effective Disease-Aware Prompting (DAP) process, which uses the explainability map of a VLM to identify the appropriate image features. This simple strategy amplifies disease-relevant regions while suppressing background interference. Without any additional pixel-level annotations, DAP improves visual grounding accuracy by 20.74% compared to state-of-the-art methods across three major chest X-ray datasets.

Tue M. Cao, Nhat X. Hoang, Hieu H. Pham et al.

Understanding the inner workings of neural networks is essential for enhancing model performance and interpretability. Current research predominantly focuses on examining the connection between individual neurons and the model's final predictions. Which suffers from challenges in interpreting the internal workings of the model, particularly when neurons encode multiple unrelated features. In this paper, we propose a novel framework that transitions the focus from analyzing individual neurons to investigating groups of neurons, shifting the emphasis from neuron-output relationships to functional interaction between neurons. Our automated framework, NeurFlow, first identifies core neurons and clusters them into groups based on shared functional relationships, enabling a more coherent and interpretable view of the network's internal processes. This approach facilitates the construction of a hierarchical circuit representing neuron interactions across layers, thus improving interpretability while reducing computational costs. Our extensive empirical studies validate the fidelity of our proposed NeurFlow. Additionally, we showcase its utility in practical applications such as image debugging and automatic concept labeling, thereby highlighting its potential to advance the field of neural network explainability.

Nam Duong Tran, Phi Le Nguyen

Recent advances in Vision-Language Models (VLMs) have greatly enhanced the integration of visual perception and linguistic reasoning, driving rapid progress in multimodal understanding. Despite these achievements, the security of VLMs, particularly their vulnerability to backdoor attacks, remains significantly underexplored. Existing backdoor attacks on VLMs are still in an early stage of development, with most current methods relying on generating poisoned responses that contain fixed, easily identifiable patterns. In this work, we make two key contributions. First, we demonstrate for the first time that the stealthiness of existing VLM backdoor attacks has been substantially overestimated. By adapting defense techniques originally designed for other domains (e.g., vision-only and text-only models), we show that several state-of-the-art attacks can be detected with surprising ease. Second, to address this gap, we introduce Phantasia, a context-adaptive backdoor attack that dynamically aligns its poisoned outputs with the semantics of each input. Instead of producing static poisoned patterns, Phantasia encourages models to generate contextually coherent yet malicious responses that remain plausible, thereby significantly improving stealth and adaptability. Extensive experiments across diverse VLM architectures reveal that Phantasia achieves state-of-the-art attack success rates while maintaining benign performance under various defensive settings.

Tuan Dung Nguyen, Minh Khoi Ho, Qi Chen et al.

Large vision-language models (LVLMs) achieve strong performance on visual reasoning tasks but remain highly susceptible to hallucination. Existing detection methods predominantly rely on coarse, whole-image measures of how an object token relates to the input image. This global strategy is limited: hallucinated tokens may exhibit weak but widely scattered correlations across many local regions, which aggregate into deceptively high overall relevance, thus evading the current global hallucination detectors. We begin with a simple yet critical observation: a faithful object token must be strongly grounded in a specific image region. Building on this insight, we introduce a patch-level hallucination detection framework that examines fine-grained token-level interactions across model layers. Our analysis uncovers two characteristic signatures of hallucinated tokens: (i) they yield diffuse, non-localized attention patterns, in contrast to the compact, well-focused attention seen in faithful tokens; and (ii) they fail to exhibit meaningful semantic alignment with any visual region. Guided by these findings, we develop a lightweight and interpretable detection method that leverages patch-level statistical features, combined with hidden-layer representations. Our approach achieves up to 90% accuracy in token-level hallucination detection, demonstrating the superiority of fine-grained structural analysis for detecting hallucinations.

Manh Duong Nguyen, Dac Thai Nguyen, Trung Viet Nguyen et al.

Osteosarcoma, the most common primary bone cancer, often requires accurate necrosis assessment from whole slide images (WSIs) for effective treatment planning and prognosis. However, manual assessments are subjective and prone to variability. In response, we introduce FDDM, a novel framework bridging the gap between patch classification and region-based segmentation. FDDM operates in two stages: patch-based classification, followed by region-based refinement, enabling cross-patch information intergation. Leveraging a newly curated dataset of osteosarcoma images, FDDM demonstrates superior segmentation performance, achieving up to a 10% improvement mIOU and a 32.12% enhancement in necrosis rate estimation over state-of-the-art methods. This framework sets a new benchmark in osteosarcoma assessment, highlighting the potential of foundation models and diffusion-based refinements in complex medical imaging tasks.

Abin Shoby, Ta Duc Huy, Tuan Dung Nguyen et al.

Vision Language models (VLMs) often hallucinate non-existent objects. Detecting hallucination is analogous to detecting deception: a single final statement is insufficient, one must examine the underlying reasoning process. Yet existing detectors rely mostly on final-layer signals. Attention-based methods assume hallucinated tokens exhibit low attention, while entropy-based ones use final-step uncertainty. Our analysis reveals the opposite: hallucinated objects can exhibit peaked attention due to contextual priors; and models often express high confidence because intermediate layers have already converged to an incorrect hypothesis. We show that the key to hallucination detection lies within the model's thought process, not its final output. By probing decoder layers, we uncover a previously overlooked behavior, overthinking: models repeatedly revise object hypotheses across layers before committing to an incorrect answer. Once the model latches onto a confounded hypothesis, it can propagate through subsequent layers, ultimately causing hallucination. To capture this behavior, we introduce the Overthinking Score, a metric to measure how many competing hypotheses the model entertains and how unstable these hypotheses are across layers. This score significantly improves hallucination detection: 78.9% F1 on MSCOCO and 71.58% on AMBER.

Tuyen Nguyen, Trong Nghia Hoang, Phi Le Nguyen et al.

The aim of this paper is to introduce a quantum fusion mechanism for multimodal learning and to establish its theoretical and empirical potential. The proposed method, called the Quantum Fusion Layer (QFL), replaces classical fusion schemes with a hybrid quantum-classical procedure that uses parameterized quantum circuits to learn entangled feature interactions without requiring exponential parameter growth. Supported by quantum signal processing principles, the quantum component efficiently represents high-order polynomial interactions across modalities with linear parameter scaling, and we provide a separation example between QFL and low-rank tensor-based methods that highlights potential quantum query advantages. In simulation, QFL consistently outperforms strong classical baselines on small but diverse multimodal tasks, with particularly marked improvements in high-modality regimes. These results suggest that QFL offers a fundamentally new and scalable approach to multimodal fusion that merits deeper exploration on larger systems.

Nam Duong Tran, Nam Nguyen Phuong, Hieu H. Pham et al.

Deep neural networks often suffer performance drops when test data distribution differs from training data. Domain Generalization (DG) aims to address this by focusing on domain-invariant features or augmenting data for greater diversity. However, these methods often struggle with limited training domains or significant gaps between seen (training) and unseen (test) domains. To enhance DG robustness, we hypothesize that it is essential for the model to be trained on data from domains that closely resemble unseen test domains-an inherently difficult task due to the absence of prior knowledge about the unseen domains. Accordingly, we propose ConstStyle, a novel approach that leverages a unified domain to capture domain-invariant features and bridge the domain gap with theoretical analysis. During training, all samples are mapped onto this unified domain, optimized for seen domains. During testing, unseen domain samples are projected similarly before predictions. By aligning both training and testing data within this unified domain, ConstStyle effectively reduces the impact of domain shifts, even with large domain gaps or few seen domains. Extensive experiments demonstrate that ConstStyle consistently outperforms existing methods across diverse scenarios. Notably, when only a limited number of seen domains are available, ConstStyle can boost accuracy up to 19.82\% compared to the next best approach.

Bao Hieu Tran, Thanh Le-Cong, Huu Manh Nguyen et al.

In the last decades, scene text recognition has gained worldwide attention from both the academic community and actual users due to its importance in a wide range of applications. Despite achievements in optical character recognition, scene text recognition remains challenging due to inherent problems such as distortions or irregular layout. Most of the existing approaches mainly leverage recurrence or convolution-based neural networks. However, while recurrent neural networks (RNNs) usually suffer from slow training speed due to sequential computation and encounter problems as vanishing gradient or bottleneck, CNN endures a trade-off between complexity and performance. In this paper, we introduce SAFL, a self-attention-based neural network model with the focal loss for scene text recognition, to overcome the limitation of the existing approaches. The use of focal loss instead of negative log-likelihood helps the model focus more on low-frequency samples training. Moreover, to deal with the distortions and irregular texts, we exploit Spatial TransformerNetwork (STN) to rectify text before passing to the recognition network. We perform experiments to compare the performance of the proposed model with seven benchmarks. The numerical results show that our model achieves the best performance.