Vinh Nguyen

Khang Nguyen, Hieu Nong, Vinh Nguyen et al.

Graph Neural Networks (GNNs) had been demonstrated to be inherently susceptible to the problems of over-smoothing and over-squashing. These issues prohibit the ability of GNNs to model complex graph interactions by limiting their effectiveness in taking into account distant information. Our study reveals the key connection between the local graph geometry and the occurrence of both of these issues, thereby providing a unified framework for studying them at a local scale using the Ollivier-Ricci curvature. Specifically, we demonstrate that over-smoothing is linked to positive graph curvature while over-squashing is linked to negative graph curvature. Based on our theory, we propose the Batch Ollivier-Ricci Flow, a novel rewiring algorithm capable of simultaneously addressing both over-smoothing and over-squashing.

Bernal Jimenez Gutierrez, Yuqing Mao, Vinh Nguyen et al.

As the immense opportunities enabled by large language models become more apparent, NLP systems will be increasingly expected to excel in real-world settings. However, in many instances, powerful models alone will not yield translational NLP solutions, especially if the formulated problem is not well aligned with the real-world task. In this work, we study the case of UMLS vocabulary insertion, an important real-world task in which hundreds of thousands of new terms, referred to as atoms, are added to the UMLS, one of the most comprehensive open-source biomedical knowledge bases. Previous work aimed to develop an automated NLP system to make this time-consuming, costly, and error-prone task more efficient. Nevertheless, practical progress in this direction has been difficult to achieve due to a problem formulation and evaluation gap between research output and the real-world task. In order to address this gap, we introduce a new formulation for UMLS vocabulary insertion which mirrors the real-world task, datasets which faithfully represent it and several strong baselines we developed through re-purposing existing solutions. Additionally, we propose an effective rule-enhanced biomedical language model which enables important new model behavior, outperforms all strong baselines and provides measurable qualitative improvements to editors who carry out the UVI task. We hope this case study provides insight into the considerable importance of problem formulation for the success of translational NLP solutions.

Tri-Tin Nguyen, Tien-Dat Nguyen, Gia-Uy Le et al.

This paper presents a learning-based navigation framework for indoor mobile robots. The proposed method combines a supervised neural global planner, trained from cost-aware A* expert trajectories, with the proposed Learning-Based DWA local planner, which is formulated as discrete candidate selection over the Dynamic Window Approach (DWA) action lattice. For local planning, the policy is first trained by behavior cloning and then refined by Proximal Policy Optimization (PPO) under feasibility-aware masking. The framework is implemented and evaluated in both simulated and real-world indoor environments. Experimental results show that the proposed method generates feasible global routes and reliable local motion commands for safe goal-directed navigation in the presence of obstacles. These results demonstrate the effectiveness of integrating learning-based global planning with reinforcement-learning-refined local control for indoor mobile robot navigation. The source code will be released at https://ntdathp.github.io/rl_robot_web/.

Thilini Wijesiriwardene, Vinh Nguyen, Goonmeet Bajaj et al.

The UMLS Metathesaurus integrates more than 200 biomedical source vocabularies. During the Metathesaurus construction process, synonymous terms are clustered into concepts by human editors, assisted by lexical similarity algorithms. This process is error-prone and time-consuming. Recently, a deep learning model (LexLM) has been developed for the UMLS Vocabulary Alignment (UVA) task. This work introduces UBERT, a BERT-based language model, pretrained on UMLS terms via a supervised Synonymy Prediction (SP) task replacing the original Next Sentence Prediction (NSP) task. The effectiveness of UBERT for UMLS Metathesaurus construction process is evaluated using the UMLS Vocabulary Alignment (UVA) task. We show that UBERT outperforms the LexLM, as well as biomedical BERT-based models. Key to the performance of UBERT are the synonymy prediction task specifically developed for UBERT, the tight alignment of training data to the UVA task, and the similarity of the models used for pretrained UBERT.

Vinh Nguyen, Gia-Uy Le, Tien-Dat Nguyen et al.

With the rapid development of simulation tools, the development and validation of autonomous robotic systems have become more efficient before real-world deployment. This paper presents a simulation-to-real implementation of an autonomous mobile robot based on an existing mechanical platform. Instead of focusing on mechanical design, our work concentrates on the development of the onboard control, self-localization, and autonomous navigation system. The proposed robot is equipped with onboard sensing and computation to estimate its pose and navigate autonomously in the environment. The overall framework is first developed and tested in simulation, and then deployed on the real robot for experimental evaluation. The results demonstrate the feasibility of the proposed approach and show that simulation provides an effective foundation for developing reliable autonomous mobile robot systems. The source code will be released at https://ntdathp.github.io/outdoor-robot-web.

Vinh Nguyen, Olivier Bodenreider

The construction and maintenance process of the UMLS (Unified Medical Language System) Metathesaurus is time-consuming, costly, and error-prone as it relies on (1) the lexical and semantic processing for suggesting synonymous terms, and (2) the expertise of UMLS editors for curating the suggestions. For improving the UMLS Metathesaurus construction process, our research group has defined a new task called UVA (UMLS Vocabulary Alignment) and generated a dataset for evaluating the task. Our group has also developed different baselines for this task using logical rules (RBA), and neural networks (LexLM and ConLM). In this paper, we present a set of reusable and reproducible resources including (1) a dataset generator, (2) three datasets generated by using the generator, and (3) three baseline approaches. We describe the UVA dataset generator and its implementation generalized for any given UMLS release. We demonstrate the use of the dataset generator by generating datasets corresponding to three UMLS releases, 2020AA, 2021AA, and 2021AB. We provide three UVA baselines using the three existing approaches (LexLM, ConLM, and RBA). The code, the datasets, and the experiments are publicly available, reusable, and reproducible with any UMLS release (a no-cost license agreement is required for downloading the UMLS).

Vinh Nguyen, Cuong Dang, Jiahao Zhang et al.

Retrieval-Augmented Generation (RAG) has emerged as a widely adopted approach for enhancing LLMs in scenarios that demand extensive factual knowledge. However, current RAG evaluations concentrate primarily on correctness, which may not fully capture the impact of retrieval on LLM uncertainty and reliability. To bridge this gap, we introduce URAG, a comprehensive benchmark designed to assess the uncertainty of RAG systems across various fields like healthcare, programming, science, math, and general text. By reformulating open-ended generation tasks into multiple-choice question answering, URAG allows for principled uncertainty quantification via conformal prediction. We apply the evaluation pipeline to 8 standard RAG methods, measuring their performance through both accuracy and prediction-set sizes based on LAC and APS metrics. Our analysis shows that (1) accuracy gains often coincide with reduced uncertainty, but this relationship breaks under retrieval noise; (2) simple modular RAG methods tend to offer better accuracy-uncertainty trade-offs than more complex reasoning pipelines; and (3) no single RAG approach is universally reliable across domains. We further show that (4) retrieval depth, parametric knowledge dependence, and exposure to confidence cues can amplify confident errors and hallucinations. Ultimately, URAG establishes a systematic benchmark for analyzing and enhancing the trustworthiness of retrieval-augmented systems. Our code is available on GitHub.

Akhiad Bercovich, Itay Levy, Izik Golan et al. · nvidia

We introduce the Llama-Nemotron series of models, an open family of heterogeneous reasoning models that deliver exceptional reasoning capabilities, inference efficiency, and an open license for enterprise use. The family comes in three sizes -- Nano (8B), Super (49B), and Ultra (253B) -- and performs competitively with state-of-the-art reasoning models such as DeepSeek-R1 while offering superior inference throughput and memory efficiency. In this report, we discuss the training procedure for these models, which entails using neural architecture search from Llama 3 models for accelerated inference, knowledge distillation, and continued pretraining, followed by a reasoning-focused post-training stage consisting of two main parts: supervised fine-tuning and large scale reinforcement learning. Llama-Nemotron models are the first open-source models to support a dynamic reasoning toggle, allowing users to switch between standard chat and reasoning modes during inference. To further support open research and facilitate model development, we provide the following resources: 1. We release the Llama-Nemotron reasoning models -- LN-Nano, LN-Super, and LN-Ultra -- under the commercially permissive NVIDIA Open Model License Agreement. 2. We release the complete post-training dataset: Llama-Nemotron-Post-Training-Dataset. 3. We also release our training codebases: NeMo, NeMo-Aligner, and Megatron-LM.

Chuong H. Nguyen, Su Huynh, Vinh Nguyen et al.

Since being introduced in 2020, Vision Transformers (ViT) has been steadily breaking the record for many vision tasks and are often described as ``all-you-need" to replace ConvNet. Despite that, ViTs are generally computational, memory-consuming, and unfriendly for embedded devices. In addition, recent research shows that standard ConvNet if redesigned and trained appropriately can compete favorably with ViT in terms of accuracy and scalability. In this paper, we adopt the modernized structure of ConvNet to design a new backbone for action recognition. Particularly, our main target is to serve for industrial product deployment, such as FPGA boards in which only standard operations are supported. Therefore, our network simply consists of 2D convolutions, without using any 3D convolution, long-range attention plugin, or Transformer blocks. While being trained with much fewer epochs (5x-10x), our backbone surpasses the methods using (2+1)D and 3D convolution, and achieve comparable results with ViT on two benchmark datasets.

Tuan Nguyen, Hirotada Honda, Takashi Sano et al.

We propose the Kuramoto Graph Neural Network (KuramotoGNN), a novel class of continuous-depth graph neural networks (GNNs) that employs the Kuramoto model to mitigate the over-smoothing phenomenon, in which node features in GNNs become indistinguishable as the number of layers increases. The Kuramoto model captures the synchronization behavior of non-linear coupled oscillators. Under the view of coupled oscillators, we first show the connection between Kuramoto model and basic GNN and then over-smoothing phenomenon in GNNs can be interpreted as phase synchronization in Kuramoto model. The KuramotoGNN replaces this phase synchronization with frequency synchronization to prevent the node features from converging into each other while allowing the system to reach a stable synchronized state. We experimentally verify the advantages of the KuramotoGNN over the baseline GNNs and existing methods in reducing over-smoothing on various graph deep learning benchmark tasks.

Tuan Nguyen, Tam Nguyen, Vinh Nguyen et al.

$p$-Laplacian regularization, rooted in graph and image signal processing, introduces a parameter $p$ to control the regularization effect on these data. Smaller values of $p$ promote sparsity and interpretability, while larger values encourage smoother solutions. In this paper, we first show that the self-attention mechanism obtains the minimal Laplacian regularization ($p=2$) and encourages the smoothness in the architecture. However, the smoothness is not suitable for the heterophilic structure of self-attention in transformers where attention weights between tokens that are in close proximity and non-close ones are assigned indistinguishably. From that insight, we then propose a novel class of transformers, namely the $p$-Laplacian Transformer (p-LaT), which leverages $p$-Laplacian regularization framework to harness the heterophilic features within self-attention layers. In particular, low $p$ values will effectively assign higher attention weights to tokens that are in close proximity to the current token being processed. We empirically demonstrate the advantages of p-LaT over the baseline transformers on a wide range of benchmark datasets.

Yasaman Hashem Pour, Nazanin Mahjourian, Vinh Nguyen

Manual generation of G-code is important for learning the operation of CNC machines. Prior work in G-code verification uses Large-Language Models (LLMs), which primarily examine errors in the written programming. However, CNC machining requires extensive use and knowledge of the Human-Machine Interface (HMI), which displays machine status and errors. LLMs currently lack the capability to leverage knowledge of HMIs due to their inability to access the vision modality. This paper proposes a few-shot VLM-based verification approach that simultaneously evaluates the G-code and the HMI display for errors and safety status. The input dataset includes paired G-code text and associated HMI screenshots from a 15-slant-PRO lathe, including both correct and error-prone cases. To enable few-shot learning, the VLM is provided with a structured JSON schema based on prior heuristic knowledge. After determining the prompts, instances of G-code and HMI that either contain errors or are error free are used as few-shot examples to guide the VLM. The model was then evaluated in comparison to a zero-shot VLM through multiple scenarios of incorrect G-code and HMI errors with respect to per-slot accuracy. The VLM showed that few-shot prompting led to overall enhancement of detecting HMI errors and discrepancies with the G-code for more comprehensive debugging. Therefore, the proposed framework was demonstrated to be suitable for verification of manually generated G-code that is typically developed in CNC training.

Nazanin Mahjourian, Vinh Nguyen

Manufacturing requires reliable object detection methods for precise picking and handling of diverse types of manufacturing parts and components. Traditional object detection methods utilize either only 2D images from cameras or 3D data from lidars or similar 3D sensors. However, each of these sensors have weaknesses and limitations. Cameras do not have depth perception and 3D sensors typically do not carry color information. These weaknesses can undermine the reliability and robustness of industrial manufacturing systems. To address these challenges, this work proposes a multi-sensor system combining an red-green-blue (RGB) camera and a 3D point cloud sensor. The two sensors are calibrated for precise alignment of the multimodal data captured from the two hardware devices. A novel multimodal object detection method is developed to process both RGB and depth data. This object detector is based on the Faster R-CNN baseline that was originally designed to process only camera images. The results show that the multimodal model significantly outperforms the depth-only and RGB-only baselines on established object detection metrics. More specifically, the multimodal model improves mAP by 13% and raises Mean Precision by 11.8% in comparison to the RGB-only baseline. Compared to the depth-only baseline, it improves mAP by 78% and raises Mean Precision by 57%. Hence, this method facilitates more reliable and robust object detection in service to smart manufacturing applications.

Nazanin Mahjourian, Vinh Nguyen

The success of machine learning models in industrial applications is heavily dependent on the quality of the datasets used to train the models. However, large-scale datasets, specially those constructed from crowd-sourcing and web-scraping, often suffer from label noise, inconsistencies, and errors. This problem is particularly pronounced in manufacturing domains, where obtaining high-quality labels is costly and time-consuming. This paper introduces Vision-Language Sanitization and Refinement (VLSR), which is a vision-language-based framework for label sanitization and refinement in multi-label manufacturing image datasets. This method embeds both images and their associated textual labels into a shared semantic space leveraging the CLIP vision-language model. Then two key tasks are addressed in this process by computing the cosine similarity between embeddings. First, label sanitization is performed to identify irrelevant, misspelled, or semantically weak labels, and surface the most semantically aligned label for each image by comparing image-label pairs using cosine similarity between image and label embeddings. Second, the method applies density-based clustering on text embeddings, followed by iterative cluster merging, to group semantically similar labels into unified label groups. The Factorynet dataset, which includes noisy labels from both human annotations and web-scraped sources, is employed to evaluate the effectiveness of the proposed framework. Experimental results demonstrate that the VLSR framework successfully identifies problematic labels and improves label consistency. This method enables a significant reduction in label vocabulary through clustering, which ultimately enhances the dataset's quality for training robust machine learning models in industrial applications with minimal human intervention.

Nazanin Mahjourian, Vinh Nguyen

Many manufacturing environments operate in low-light conditions or within enclosed machines where conventional vision systems struggle. Infrared cameras provide complementary advantages in such environments. Simultaneously, supervised AI systems require large labeled datasets, which makes zero-shot learning frameworks more practical for applications including infrared cameras. Recent advances in vision-language foundation models (VLMs) offer a new path in zero-shot predictions from paired image-text representations. However, current VLMs cannot understand infrared camera data since they are trained on RGB data. This work introduces VLM-IRIS (Vision-Language Models for InfraRed Industrial Sensing), a zero-shot framework that adapts VLMs to infrared data by preprocessing infrared images captured by a FLIR Boson sensor into RGB-compatible inputs suitable for CLIP-based encoders. We demonstrate zero-shot workpiece presence detection on a 3D printer bed where temperature differences between the build plate and workpieces make the task well-suited for thermal imaging. VLM-IRIS converts the infrared images to magma representation and applies centroid prompt ensembling with a CLIP ViT-B/32 encoder to achieve high accuracy on infrared images without any model retraining. These findings demonstrate that the proposed improvements to VLMs can be effectively extended to thermal applications for label-free monitoring.

Vinh Nguyen

Generating detailed descriptions from multiple cameras and viewpoints is challenging due to the complex and inconsistent nature of visual data. In this paper, we introduce PerspectiveNet, a lightweight yet efficient model for generating long descriptions across multiple camera views. Our approach utilizes a vision encoder, a compact connector module to convert visual features into a fixed-size tensor, and large language models (LLMs) to harness the strong natural language generation capabilities of LLMs. The connector module is designed with three main goals: mapping visual features onto LLM embeddings, emphasizing key information needed for description generation, and producing a fixed-size feature matrix. Additionally, we augment our solution with a secondary task, the correct frame sequence detection, enabling the model to search for the correct sequence of frames to generate descriptions. Finally, we integrate the connector module, the secondary task, the LLM, and a visual feature extraction model into a single architecture, which is trained for the Traffic Safety Description and Analysis task. This task requires generating detailed, fine-grained descriptions of events from multiple cameras and viewpoints. The resulting model is lightweight, ensuring efficient training and inference, while remaining highly effective.

Goonmeet Bajaj, Vinh Nguyen, Thilini Wijesiriwardene et al.

The current UMLS (Unified Medical Language System) Metathesaurus construction process for integrating over 200 biomedical source vocabularies is expensive and error-prone as it relies on the lexical algorithms and human editors for deciding if the two biomedical terms are synonymous. Recent advances in Natural Language Processing such as Transformer models like BERT and its biomedical variants with contextualized word embeddings have achieved state-of-the-art (SOTA) performance on downstream tasks. We aim to validate if these approaches using the BERT models can actually outperform the existing approaches for predicting synonymy in the UMLS Metathesaurus. In the existing Siamese Networks with LSTM and BioWordVec embeddings, we replace the BioWordVec embeddings with the biomedical BERT embeddings extracted from each BERT model using different ways of extraction. In the Transformer architecture, we evaluate the use of the different biomedical BERT models that have been pre-trained using different datasets and tasks. Given the SOTA performance of these BERT models for other downstream tasks, our experiments yield surprisingly interesting results: (1) in both model architectures, the approaches employing these biomedical BERT-based models do not outperform the existing approaches using Siamese Network with BioWordVec embeddings for the UMLS synonymy prediction task, (2) the original BioBERT large model that has not been pre-trained with the UMLS outperforms the SapBERT models that have been pre-trained with the UMLS, and (3) using the Siamese Networks yields better performance for synonymy prediction when compared to using the biomedical BERT models.

Oscar Correa, Jeffrey Chan, Vinh Nguyen

Many approaches have been proposed to discover clusters within networks. Community finding field encompasses approaches which try to discover clusters where nodes are tightly related within them but loosely related with nodes of other clusters. However, a community network configuration is not the only possible latent structure in a graph. Core-periphery and hierarchical network configurations are valid structures to discover in a relational dataset. On the other hand, a network is not completely explained by only knowing the membership of each node. A high level view of the inter-cluster relationships is needed. Blockmodelling techniques deal with these two issues. Firstly, blockmodelling allows finding any network configuration besides to the well-known community structure. Secondly, blockmodelling is a summary representation of a network which regards not only membership of nodes but also relations between clusters. Finally, a unique summary representation of a network is unlikely. Networks might hide more than one blockmodel. Therefore, our proposed problem aims to discover a secondary blockmodel representation of a network that is of good quality and dissimilar with respect to a given blockmodel. Our methodology is presented through two approaches, (a) inclusion of cannot-link constraints and (b) dissimilarity between image matrices. Both approaches are based on non-negative matrix factorisation NMF which fits the blockmodelling representation. The evaluation of these two approaches regards quality and dissimilarity of the discovered alternative blockmodel as these are the requirements of the problem.

Vinh Nguyen, Md Yasin Kabir, Tommy Dang

Interactive visualization tools are highly desirable to biologist and cancer researchers to explore the complex structures, detect patterns and find out the relationships among bio-molecules responsible for a cancer type. A pathway contains various bio-molecules in different layers of the cell which is responsible for specific cancer type. Researchers are highly interested in understanding the relationships among the proteins of different pathways and furthermore want to know how those proteins are interacting in different pathways for various cancer types. Biologists find it useful to merge the data of different cancer studies in a single network and see the relationships among the different proteins which can help them detect the common proteins in cancer studies and hence reveal the pattern of interactions of those proteins. We introduce the CancerLinker, a visual analytic tool that helps researchers explore cancer study interaction network. Twenty-six cancer studies are merged to explore pathway data and bio-molecules relationships that can provide the answers to some significant questions which are helpful in cancer research. The CancerLinker also helps biologists explore the critical mutated proteins in multiple cancer studies. A bubble graph is constructed to visualize common protein based on its frequency and biological assemblies. Parallel coordinates highlight patterns of patient profiles (obtained from cBioportal by WebAPI services) on different attributes for a specified cancer study

Vinh Nguyen, Amit Sheth

Logical inference, an integral feature of the Semantic Web, is the process of deriving new triples by applying entailment rules on knowledge bases. The entailment rules are determined by the model-theoretic semantics. Incorporating context of an RDF triple (e.g., provenance, time, and location) into the inferencing process requires the formal semantics to be capable of describing the context of RDF triples also in the form of triples, or in other words, RDF contextual triples about triples. The formal semantics should also provide the rules that could entail new contextual triples about triples. In this paper, we propose the first inferencing mechanism that allows context of RDF triples, represented in the form of RDF triples about triples, to be the first-class citizens in the model-theoretic semantics and in the logical rules. Our inference mechanism is well-formalized with all new concepts being captured in the model-theoretic semantics. This formal semantics also allows us to derive a new set of entailment rules that could entail new contextual triples about triples. To demonstrate the feasibility and the scalability of the proposed mechanism, we implement a new tool in which we transform the existing knowledge bases to our representation of RDF triples about triples and provide the option for this tool to compute the inferred triples for the proposed rules. We evaluate the computation of the proposed rules on a large scale using various real-world knowledge bases such as Bio2RDF NCBI Genes and DBpedia. The results show that the computation of the inferred triples can be highly scalable. On average, one billion inferred triples adds 5-6 minutes to the overall transformation process. NCBI Genes, with 20 billion triples in total, took only 232 minutes for the transformation of 12 billion triples and added 42 minutes for inferring 8 billion triples to the overall process.

Vinh Nguyen, Olivier Bodenreider, Krishnaprasad Thirunarayan et al.

The Singleton Property (SP) approach has been proposed for representing and querying metadata about RDF triples such as provenance, time, location, and evidence. In this approach, one singleton property is created to uniquely represent a relationship in a particular context, and in general, generates a large property hierarchy in the schema. It has become the subject of important questions from Semantic Web practitioners. Can an existing reasoner recognize the singleton property triples? And how? If the singleton property triples describe a data triple, then how can a reasoner infer this data triple from the singleton property triples? Or would the large property hierarchy affect the reasoners in some way? We address these questions in this paper and present our study about the reasoning aspects of the singleton properties. We propose a simple mechanism to enable existing reasoners to recognize the singleton property triples, as well as to infer the data triples described by the singleton property triples. We evaluate the effect of the singleton property triples in the reasoning processes by comparing the performance on RDF datasets with and without singleton properties. Our evaluation uses as benchmark the LUBM datasets and the LUBM-SP datasets derived from LUBM with temporal information added through singleton properties.