Showing 1–50 of 116 results for author: Vu, D

The Scaffold Effect: How Prompt Framing Drives Apparent Multimodal Gains in Clinical VLM Evaluation

Authors: Doan Nam Long Vu, Simone Balloccu

Abstract: Trustworthy clinical AI requires that performance gains reflect genuine evidence integration rather than surface-level artifacts. We evaluate 12 open-weight vision-language models (VLMs) on binary classification across two clinical neuroimaging cohorts, \textsc{FOR2107} (affective disorders) and \textsc{OASIS-3} (cognitive decline). Both datasets come with structural MRI data that carries no relia… ▽ More Trustworthy clinical AI requires that performance gains reflect genuine evidence integration rather than surface-level artifacts. We evaluate 12 open-weight vision-language models (VLMs) on binary classification across two clinical neuroimaging cohorts, \textsc{FOR2107} (affective disorders) and \textsc{OASIS-3} (cognitive decline). Both datasets come with structural MRI data that carries no reliable individual-level diagnostic signal. Under these conditions, smaller VLMs exhibit gains of up to 58\% F1 upon introduction of neuroimaging context, with distilled models becoming competitive with counterparts an order of magnitude larger. A contrastive confidence analysis reveals that merely \emph{mentioning} MRI availability in the task prompt accounts for 70-80\% of this shift, independent of whether imaging data is present, a domain-specific instance of modality collapse we term the \emph{scaffold effect}. Expert evaluation reveals fabrication of neuroimaging-grounded justifications across all conditions, and preference alignment, while eliminating MRI-referencing behavior, collapses both conditions toward random baseline. Our findings demonstrate that surface evaluations are inadequate indicators of multimodal reasoning, with direct implications for the deployment of VLMs in clinical settings. △ Less

Submitted 30 March, 2026; originally announced March 2026.

TwinMixing: A Shuffle-Aware Feature Interaction Model for Multi-Task Segmentation

Authors: Minh-Khoi Do, Huy Che, Dinh-Duy Phan, Duc-Khai Lam, Duc-Lung Vu

Abstract: Accurate and efficient perception is essential for autonomous driving, where segmentation tasks such as drivable-area and lane segmentation provide critical cues for motion planning and control. However, achieving high segmentation accuracy while maintaining real-time performance on low-cost hardware remains a challenging problem. To address this issue, we introduce TwinMixing, a lightweight multi… ▽ More Accurate and efficient perception is essential for autonomous driving, where segmentation tasks such as drivable-area and lane segmentation provide critical cues for motion planning and control. However, achieving high segmentation accuracy while maintaining real-time performance on low-cost hardware remains a challenging problem. To address this issue, we introduce TwinMixing, a lightweight multi-task segmentation model designed explicitly for drivable-area and lane segmentation. The proposed network features a shared encoder and task-specific decoders, enabling both feature sharing and task specialization. Within the encoder, we propose an Efficient Pyramid Mixing (EPM) module that enhances multi-scale feature extraction through a combination of grouped convolutions, depthwise dilated convolutions and channel shuffle operations, effectively expanding the receptive field while minimizing computational cost. Each decoder adopts a Dual-Branch Upsampling (DBU) Block composed of a learnable transposed convolution-based Fine detailed branch and a parameter-free bilinear interpolation-based Coarse grained branch, achieving detailed yet spatially consistent feature reconstruction. Extensive experiments on the BDD100K dataset validate the effectiveness of TwinMixing across three configurations - tiny, base, and large. Among them, the base configuration achieves the best trade-off between accuracy and computational efficiency, reaching 92.0% mIoU for drivable-area segmentation and 32.3% IoU for lane segmentation with only 0.43M parameters and 3.95 GFLOPs. Moreover, TwinMixing consistently outperforms existing segmentation models on the same tasks, as illustrated in Fig. 1. Thanks to its compact and modular design, TwinMixing demonstrates strong potential for real-time deployment in autonomous driving and embedded perception systems. The source code: https://github.com/Jun0se7en/TwinMixing. △ Less

Submitted 30 March, 2026; originally announced March 2026.

Journal ref: Results in Engineering 30 (2026) 109982

Anti-I2V: Safeguarding your photos from malicious image-to-video generation

Authors: Duc Vu, Anh Nguyen, Chi Tran, Anh Tran

Abstract: Advances in diffusion-based video generation models, while significantly improving human animation, poses threats of misuse through the creation of fake videos from a specific person's photo and text prompts. Recent efforts have focused on adversarial attacks that introduce crafted perturbations to protect images from diffusion-based models. However, most existing approaches target image generatio… ▽ More Advances in diffusion-based video generation models, while significantly improving human animation, poses threats of misuse through the creation of fake videos from a specific person's photo and text prompts. Recent efforts have focused on adversarial attacks that introduce crafted perturbations to protect images from diffusion-based models. However, most existing approaches target image generation, while relatively few explicitly address image-to-video diffusion models (VDMs), and most primarily focus on UNet-based architectures. Hence, their effectiveness against Diffusion Transformer (DiT) models remains largely under-explored, as these models demonstrate improved feature retention, and stronger temporal consistency due to larger capacity and advanced attention mechanisms. In this work, we introduce Anti-I2V, a novel defense against malicious human image-to-video generation, applicable across diverse diffusion backbones. Instead of restricting noise updates to the RGB space, Anti-I2V operates in both the $L$*$a$*$b$* and frequency domains, improving robustness and concentrating on salient pixels. We then identify the network layers that capture the most distinct semantic features during the denoising process to design appropriate training objectives that maximize degradation of temporal coherence and generation fidelity. Through extensive validation, Anti-I2V demonstrates state-of-the-art defense performance against diverse video diffusion models, offering an effective solution to the problem. △ Less

Submitted 25 March, 2026; originally announced March 2026.

Comments: Accepted to CVPR 2026 (Main Conference)

InverFill: One-Step Inversion for Enhanced Few-Step Diffusion Inpainting

Authors: Duc Vu, Kien Nguyen, Trong-Tung Nguyen, Ngan Nguyen, Phong Nguyen, Khoi Nguyen, Cuong Pham, Anh Tran

Abstract: Recent diffusion-based models achieve photorealism in image inpainting but require many sampling steps, limiting practical use. Few-step text-to-image models offer faster generation, but naively applying them to inpainting yields poor harmonization and artifacts between the background and inpainted region. We trace this cause to random Gaussian noise initialization, which under low function evalua… ▽ More Recent diffusion-based models achieve photorealism in image inpainting but require many sampling steps, limiting practical use. Few-step text-to-image models offer faster generation, but naively applying them to inpainting yields poor harmonization and artifacts between the background and inpainted region. We trace this cause to random Gaussian noise initialization, which under low function evaluations causes semantic misalignment and reduced fidelity. To overcome this, we propose InverFill, a one-step inversion method tailored for inpainting that injects semantic information from the input masked image into the initial noise, enabling high-fidelity few-step inpainting. Instead of training inpainting models, InverFill leverages few-step text-to-image models in a blended sampling pipeline with semantically aligned noise as input, significantly improving vanilla blended sampling and even matching specialized inpainting models at low NFEs. Moreover, InverFill does not require real-image supervision and only adds minimal inference overhead. Extensive experiments show that InverFill consistently boosts baseline few-step models, improving image quality and text coherence without costly retraining or heavy iterative optimization. △ Less

Submitted 24 March, 2026; originally announced March 2026.

Comments: Accepted to CVPR'26 (Main Conference)

Hierarchical Decoding for Discrete Speech Synthesis with Multi-Resolution Spoof Detection

Authors: Junchuan Zhao, Minh Duc Vu, Ye Wang

Abstract: Neural codec language models enable high-quality discrete speech synthesis, yet their inference remains vulnerable to token-level artifacts and distributional drift that degrade perceptual realism. Rather than relying on preference optimization or retraining, we propose MSpoof-TTS, a training-free inference framework that improves zero-shot synthesis through multi-resolution spoof guidance. We int… ▽ More Neural codec language models enable high-quality discrete speech synthesis, yet their inference remains vulnerable to token-level artifacts and distributional drift that degrade perceptual realism. Rather than relying on preference optimization or retraining, we propose MSpoof-TTS, a training-free inference framework that improves zero-shot synthesis through multi-resolution spoof guidance. We introduce a Multi-Resolution Token-based Spoof Detection framework that evaluates codec sequences at different temporal granularities to detect locally inconsistent or unnatural patterns. We then integrate the spoof detectors into a hierarchical decoding strategy, progressively pruning low-quality candidates and re-ranking hypotheses. This discriminator-guided generation enhances robustness without modifying model parameters. Experiments validate the effectiveness of our framework for robust and high-quality codec-based speech generation. Audio samples are available at https://danny-nus.github.io/MSpoofTTS.github.io/. △ Less

Submitted 11 April, 2026; v1 submitted 5 March, 2026; originally announced March 2026.

Comments: 7 pages, 3 figures, 3 tables, 2 algorithms

DWA-KD: Dual-Space Weighting and Time-Warped Alignment for Cross-Tokenizer Knowledge Distillation

Authors: Duc Trung Vu, Pham Khanh Chi, Dat Phi Van, Linh Ngo Van, Sang Dinh, Trung Le

Abstract: Knowledge Distillation (KD) has emerged as a crucial technique for compressing Large Language Models (LLMs). Although existing cross-tokenizer KD methods have made notable progress, their effectiveness remains constrained by suboptimal alignment across sequence and vocabulary levels. To address these limitations, we introduce Dual-Space Weighting and Time-Warped Alignment (DWA-KD), a novel cross-t… ▽ More Knowledge Distillation (KD) has emerged as a crucial technique for compressing Large Language Models (LLMs). Although existing cross-tokenizer KD methods have made notable progress, their effectiveness remains constrained by suboptimal alignment across sequence and vocabulary levels. To address these limitations, we introduce Dual-Space Weighting and Time-Warped Alignment (DWA-KD), a novel cross-tokenizer distillation framework that enhances token-wise distillation through dual-space entropy-based weighting and achieves precise sequence-level alignment by leveraging both lexical and semantic information. At the token level, DWA-KD maps teacher representations into the student space and vice versa, performing dual-space KD via Kullback-Leibler divergence (KL). The process is modulated by dual-space weights that up-weight tokens where the student is uncertain and the teacher is confident, thereby focusing learning on informative tokens rather than treating all positions equally. At the sequence level, DWA-KD applies Soft Dynamic Time Warping (Soft-DTW) to both the embedding and final hidden-state layers, enabling robust alignment of lexical and contextual semantics between teacher and student sequences. Extensive experiments across diverse NLP benchmarks demonstrate that DWA-KD outperforms state-of-the-art KD baselines, while ablation studies confirm the complementary contributions of entropy-based token weighting and embedding and final hidden state layer Soft-DTW alignment. △ Less

Submitted 25 February, 2026; originally announced February 2026.

Comments: EACL Findings

Potential Role of Agentic Artificial Intelligence in Toxicologic Pathology

Authors: Nasir Rajpoot, Richard Haworth, Xavier Palazzi, Alok Sharma, Manu Sebastian, Stephen Cahalan, Dinesh S. Bangari, Radhakrishna Sura, James Hartke, Marco Tecilla, Krishna Yekkala, Simon Graham, Dang Vu, David Snead, Mostafa Jahanifar, Adnan Khan, Erio Barale-Thomas

Abstract: As the volume and complexity of nonclinical toxicology studies continue to increase, toxicologic pathology reporting faces persistent challenges, including fragmented sources of data (e.g., histopathology images, clinical pathology and other study data, adverse effects database, mechanistic literature), variable reporting timelines and heightened regulatory expectations. This white paper examines… ▽ More As the volume and complexity of nonclinical toxicology studies continue to increase, toxicologic pathology reporting faces persistent challenges, including fragmented sources of data (e.g., histopathology images, clinical pathology and other study data, adverse effects database, mechanistic literature), variable reporting timelines and heightened regulatory expectations. This white paper examines the emerging role of agentic artificial intelligence (AI) in addressing these issues through coordinated workflow orchestration, data integration, and pathologist-in-the-loop report generation. Based on a closed-door roundtable held during the 2025 Society of Toxicologic Pathology (STP) Annual Meeting and follow-on discussions, this paper synthesizes the perspectives of leading toxicologic pathologists, toxicologists, and AI developers. It outlines the key pain points in current reporting workflows, identifies realistic near-term use cases for agentic AI, and describes major adoption barriers including requirements for transparency, validation, and organizational readiness. A phased adoption roadmap and pilot design considerations are proposed to help support responsible evaluation and deployment of agentic AI system in nonclinical settings. The paper concludes by emphasizing the need for coordinated efforts across pharmaceutical organizations, CROs, academia, and regulators to establish shared standards, benchmarks, and governance frameworks that will lead to safe, transparent, and trustworthy integration of AI into toxicologic science. △ Less

Submitted 13 February, 2026; v1 submitted 26 January, 2026; originally announced February 2026.

TSBOW: Traffic Surveillance Benchmark for Occluded Vehicles Under Various Weather Conditions

Authors: Ngoc Doan-Minh Huynh, Duong Nguyen-Ngoc Tran, Long Hoang Pham, Tai Huu-Phuong Tran, Hyung-Joon Jeon, Huy-Hung Nguyen, Duong Khac Vu, Hyung-Min Jeon, Son Hong Phan, Quoc Pham-Nam Ho, Chi Dai Tran, Trinh Le Ba Khanh, Jae Wook Jeon

Abstract: Global warming has intensified the frequency and severity of extreme weather events, which degrade CCTV signal and video quality while disrupting traffic flow, thereby increasing traffic accident rates. Existing datasets, often limited to light haze, rain, and snow, fail to capture extreme weather conditions. To address this gap, this study introduces the Traffic Surveillance Benchmark for Occlude… ▽ More Global warming has intensified the frequency and severity of extreme weather events, which degrade CCTV signal and video quality while disrupting traffic flow, thereby increasing traffic accident rates. Existing datasets, often limited to light haze, rain, and snow, fail to capture extreme weather conditions. To address this gap, this study introduces the Traffic Surveillance Benchmark for Occluded vehicles under various Weather conditions (TSBOW), a comprehensive dataset designed to enhance occluded vehicle detection across diverse annual weather scenarios. Comprising over 32 hours of real-world traffic data from densely populated urban areas, TSBOW includes more than 48,000 manually annotated and 3.2 million semi-labeled frames; bounding boxes spanning eight traffic participant classes from large vehicles to micromobility devices and pedestrians. We establish an object detection benchmark for TSBOW, highlighting challenges posed by occlusions and adverse weather. With its varied road types, scales, and viewpoints, TSBOW serves as a critical resource for advancing Intelligent Transportation Systems. Our findings underscore the potential of CCTV-based traffic monitoring, pave the way for new research and applications. The TSBOW dataset is publicly available at: https://github.com/SKKUAutoLab/TSBOW. △ Less

Submitted 5 February, 2026; originally announced February 2026.

Comments: This paper has been accepted by the 40th AAAI Conference on Artificial Intelligence (AAAI-26)

Mono3DV: Monocular 3D Object Detection with 3D-Aware Bipartite Matching and Variational Query DeNoising

Authors: Kiet Dang Vu, Trung Thai Tran, Kien Nguyen Do Trung, Duc Dung Nguyen

Abstract: While DETR-like architectures have demonstrated significant potential for monocular 3D object detection, they are often hindered by a critical limitation: the exclusion of 3D attributes from the bipartite matching process. This exclusion arises from the inherent ill-posed nature of 3D estimation from monocular image, which introduces instability during training. Consequently, high-quality 3D predi… ▽ More While DETR-like architectures have demonstrated significant potential for monocular 3D object detection, they are often hindered by a critical limitation: the exclusion of 3D attributes from the bipartite matching process. This exclusion arises from the inherent ill-posed nature of 3D estimation from monocular image, which introduces instability during training. Consequently, high-quality 3D predictions can be erroneously suppressed by 2D-only matching criteria, leading to suboptimal results. To address this, we propose Mono3DV, a novel Transformer-based framework. Our approach introduces three key innovations. First, we develop a 3D-Aware Bipartite Matching strategy that directly incorporates 3D geometric information into the matching cost, resolving the misalignment caused by purely 2D criteria. Second, it is important to stabilize the Bipartite Matching to resolve the instability occurring when integrating 3D attributes. Therefore, we propose 3D-DeNoising scheme in the training phase. Finally, recognizing the gradient vanishing issue associated with conventional denoising techniques, we propose a novel Variational Query DeNoising mechanism to overcome this limitation, which significantly enhances model performance. Without leveraging any external data, our method achieves state-of-the-art results on the KITTI 3D object detection benchmark. △ Less

Submitted 2 January, 2026; originally announced January 2026.

Taint-Based Code Slicing for LLMs-based Malicious NPM Package Detection

Authors: Dang-Khoa Nguyen, Gia-Thang Ho, Quang-Minh Pham, Tuyet A. Dang-Thi, Minh-Khanh Vu, Thanh-Cong Nguyen, Phat T. Tran-Truong, Duc-Ly Vu

Abstract: Software supply chain attacks targeting the npm ecosystem have become increasingly sophisticated, leveraging obfuscation and complex logic to evade traditional detection mechanisms. Recently, large language models (LLMs) have attracted significant attention for malicious code detection due to their strong capabilities in semantic code understanding. However, the practical deployment of LLMs in thi… ▽ More Software supply chain attacks targeting the npm ecosystem have become increasingly sophisticated, leveraging obfuscation and complex logic to evade traditional detection mechanisms. Recently, large language models (LLMs) have attracted significant attention for malicious code detection due to their strong capabilities in semantic code understanding. However, the practical deployment of LLMs in this domain is severely constrained by limited context windows and high computational costs. Naive approaches, such as token-based code splitting, often fragment semantic context, leading to degraded detection performance. To overcome these challenges, this paper introduces a novel LLM-based framework for malicious npm package detection that leverages code slicing techniques. A specialized taint-based slicing method tailored to the JavaScript ecosystem is proposed to recover malicious data flows. By isolating security-relevant logic from benign boilerplate code, the approach reduces the input code volume by over 99\% while preserving critical malicious behaviors. The framework is evaluated on a curated dataset comprising over \num{7000} malicious and benign npm packages. Experimental results using the DeepSeek-Coder-6.7B model demonstrate that the proposed approach achieves a detection accuracy of \num{87.04}\%, significantly outperforming a full-package baseline based on naive token splitting (\num{75.41}\%). These results indicate that semantically optimized input representations via code slicing not only mitigate the LLM context window bottleneck but also enhance reasoning precision for security analysis, providing an effective defense against evolving open-source software supply chain threats. △ Less

Submitted 10 January, 2026; v1 submitted 13 December, 2025; originally announced December 2025.

Comments: 21 pages, 1 figure, 5 tables, 2 algorithms

Physics-Informed Neural Koopman Machine for Interpretable Longitudinal Personalized Alzheimer's Disease Forecasting

Authors: Georgi Hrusanov, Duy-Thanh Vu, Duy-Cat Can, Sophie Tascedda, Margaret Ryan, Julien Bodelet, Katarzyna Koscielska, Carsten Magnus, Oliver Y. Chén

Abstract: Early forecasting of individual cognitive decline in Alzheimer's disease (AD) is central to disease evaluation and management. Despite advances, it is as of yet challenging for existing methodological frameworks to integrate multimodal data for longitudinal personalized forecasting while maintaining interpretability. To address this gap, we present the Neural Koopman Machine (NKM), a new machine l… ▽ More Early forecasting of individual cognitive decline in Alzheimer's disease (AD) is central to disease evaluation and management. Despite advances, it is as of yet challenging for existing methodological frameworks to integrate multimodal data for longitudinal personalized forecasting while maintaining interpretability. To address this gap, we present the Neural Koopman Machine (NKM), a new machine learning architecture inspired by dynamical systems and attention mechanisms, designed to forecast multiple cognitive scores simultaneously using multimodal genetic, neuroimaging, proteomic, and demographic data. NKM integrates analytical ($α$) and biological ($β$) knowledge to guide feature grouping and control the hierarchical attention mechanisms to extract relevant patterns. By implementing Fusion Group-Aware Hierarchical Attention within the Koopman operator framework, NKM transforms complex nonlinear trajectories into interpretable linear representations. To demonstrate NKM's efficacy, we applied it to study the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. Our results suggest that NKM consistently outperforms both traditional machine learning methods and deep learning models in forecasting trajectories of cognitive decline. Specifically, NKM (1) forecasts changes of multiple cognitive scores simultaneously, (2) quantifies differential biomarker contributions to predicting distinctive cognitive scores, and (3) identifies brain regions most predictive of cognitive deterioration. Together, NKM advances personalized, interpretable forecasting of future cognitive decline in AD using past multimodal data through an explainable, explicit system and reveals potential multimodal biological underpinnings of AD progression. △ Less

Submitted 5 December, 2025; originally announced December 2025.

A Trainable Centrality Framework for Modern Data

Authors: Minh Duc Vu, Mingshuo Liu, Doudou Zhou

Abstract: Measuring how central or typical a data point is underpins robust estimation, ranking, and outlier detection, but classical depth notions become expensive and unstable in high dimensions and are hard to extend beyond Euclidean data. We introduce Fused Unified centrality Score Estimation (FUSE), a neural centrality framework that operates on top of arbitrary representations. FUSE combines a global… ▽ More Measuring how central or typical a data point is underpins robust estimation, ranking, and outlier detection, but classical depth notions become expensive and unstable in high dimensions and are hard to extend beyond Euclidean data. We introduce Fused Unified centrality Score Estimation (FUSE), a neural centrality framework that operates on top of arbitrary representations. FUSE combines a global head, trained from pairwise distance-based comparisons to learn an anchor-free centrality score, with a local head, trained by denoising score matching to approximate a smoothed log-density potential. A single parameter between 0 and 1 interpolates between these calibrated signals, yielding depth-like centrality from different views via one forward pass. Across synthetic distributions, real images, time series, and text data, and standard outlier detection benchmarks, FUSE recovers meaningful classical ordering, reveals multi-scale geometric structures, and attains competitive performance with strong classical baselines while remaining simple and efficient. △ Less

Submitted 28 November, 2025; originally announced November 2025.

More Bias, Less Bias: BiasPrompting for Enhanced Multiple-Choice Question Answering

Authors: Duc Anh Vu, Thong Nguyen, Cong-Duy Nguyen, Viet Anh Nguyen, Anh Tuan Luu

Abstract: With the advancement of large language models (LLMs), their performance on multiple-choice question (MCQ) tasks has improved significantly. However, existing approaches face key limitations: answer choices are typically presented to LLMs without contextual grounding or explanation. This absence of context can lead to incomplete exploration of all possible answers, ultimately degrading the models'… ▽ More With the advancement of large language models (LLMs), their performance on multiple-choice question (MCQ) tasks has improved significantly. However, existing approaches face key limitations: answer choices are typically presented to LLMs without contextual grounding or explanation. This absence of context can lead to incomplete exploration of all possible answers, ultimately degrading the models' reasoning capabilities. To address these challenges, we introduce BiasPrompting, a novel inference framework that guides LLMs to generate and critically evaluate reasoning across all plausible answer options before reaching a final prediction. It consists of two components: first, a reasoning generation stage, where the model is prompted to produce supportive reasonings for each answer option, and then, a reasoning-guided agreement stage, where the generated reasonings are synthesized to select the most plausible answer. Through comprehensive evaluations, BiasPrompting demonstrates significant improvements in five widely used multiple-choice question answering benchmarks. Our experiments showcase that BiasPrompting enhances the reasoning capabilities of LLMs and provides a strong foundation for tackling complex and challenging questions, particularly in settings where existing methods underperform. △ Less

Submitted 25 November, 2025; originally announced November 2025.

Comments: Accepted at the 41st ACM/SIGAPP Symposium On Applied Computing (SAC 2026), Main Conference

Towards Classifying Benign And Malicious Packages Using Machine Learning

Authors: Thanh-Cong Nguyen, Ngoc-Thanh Nguyen, Van-Giau Ung, Duc-Ly Vu

Abstract: Recently, the number of malicious open-source packages in package repositories has been increasing dramatically. While major security scanners focus on identifying known Common Vulnerabilities and Exposures (CVEs) in open-source packages, there are very few studies on detecting malicious packages. Malicious open-source package detection typically requires static, dynamic analysis, or both. Dynamic… ▽ More Recently, the number of malicious open-source packages in package repositories has been increasing dramatically. While major security scanners focus on identifying known Common Vulnerabilities and Exposures (CVEs) in open-source packages, there are very few studies on detecting malicious packages. Malicious open-source package detection typically requires static, dynamic analysis, or both. Dynamic analysis is more effective as it can expose a package's behaviors at runtime. However, current dynamic analysis tools (e.g., ossf's package-analysis) lack an automatic method to differentiate malicious packages from benign packages. In this paper, we propose an approach to extract the features from dynamic analysis (e.g., executed commands) and leverage machine learning techniques to automatically classify packages as benign or malicious. Our evaluation of nearly 2000 packages on npm shows that the machine learning classifier achieves an AUC of 0.91 with a false positive rate of nearly 0%. △ Less

Submitted 18 November, 2025; originally announced November 2025.

Comments: 5 pages, 2 figures, 3 tables

Pack-A-Mal: A Malware Analysis Framework for Open-Source Packages

Authors: Duc-Ly Vu, Thanh-Cong Nguyen, Minh-Khanh Vu, Ngoc-Thanh Nguyen, Kim-Anh Do Thi

Abstract: The increasingly sophisticated environment in which attackers operate makes software security an even greater challenge in open-source projects, where malicious packages are prevalent. Static analysis tools, such as Malcontent, are highly useful but are often incapable of dealing with obfuscated malware. Such situations lead to an unreasonably high rate of false positives. This paper highlights th… ▽ More The increasingly sophisticated environment in which attackers operate makes software security an even greater challenge in open-source projects, where malicious packages are prevalent. Static analysis tools, such as Malcontent, are highly useful but are often incapable of dealing with obfuscated malware. Such situations lead to an unreasonably high rate of false positives. This paper highlights that dynamic analysis, rather than static analysis, provides greater insight but is also more resource-intensive for understanding software behaviour during execution. In this study, we enhance a dynamic analysis tool, package-analysis, to capture key runtime behaviours, including commands executed, files accessed, and network communications. This modification enables the use of container sandboxing technologies, such as gVisor, to analyse potentially malicious packages without significantly compromising the host system. △ Less

Submitted 23 January, 2026; v1 submitted 12 November, 2025; originally announced November 2025.

Comments: There was an error in the Case study: Malicious Solana web3.js Package. The actual number of downloads was 400,000 instead of 350,000

Roleplaying with Structure: Synthetic Therapist-Client Conversation Generation from Questionnaires

Authors: Doan Nam Long Vu, Rui Tan, Lena Moench, Svenja Jule Francke, Daniel Woiwod, Florian Thomas-Odenthal, Sanna Stroth, Tilo Kircher, Christiane Hermann, Udo Dannlowski, Hamidreza Jamalabadi, Shaoxiong Ji

Abstract: The development of AI for mental health is hindered by a lack of authentic therapy dialogues, due to strict privacy regulations and the fact that clinical sessions were historically rarely recorded. We present an LLM-driven pipeline that generates synthetic counseling dialogues based on structured client profiles and psychological questionnaires. Grounded on the principles of Cognitive Behavioral… ▽ More The development of AI for mental health is hindered by a lack of authentic therapy dialogues, due to strict privacy regulations and the fact that clinical sessions were historically rarely recorded. We present an LLM-driven pipeline that generates synthetic counseling dialogues based on structured client profiles and psychological questionnaires. Grounded on the principles of Cognitive Behavioral Therapy (CBT), our method creates synthetic therapeutic conversations for clinical disorders such as anxiety and depression. Our framework, SQPsych (Structured Questionnaire-based Psychotherapy), converts structured psychological input into natural language dialogues through therapist-client simulations. Due to data governance policies and privacy restrictions prohibiting the transmission of clinical questionnaire data to third-party services, previous methodologies relying on proprietary models are infeasible in our setting. We address this limitation by generating a high-quality corpus using open-weight LLMs, validated through human expert evaluation and LLM-based assessments. Our SQPsychLLM models fine-tuned on SQPsychConv achieve strong performance on counseling benchmarks, surpassing baselines in key therapeutic skills. Our findings highlight the potential of synthetic data to enable scalable, data-secure, and clinically informed AI for mental health support. We will release our code, models, and corpus at https://ai-mh.github.io/SQPsych △ Less

Submitted 29 October, 2025; originally announced October 2025.

Improved Training Technique for Shortcut Models

Authors: Anh Nguyen, Viet Nguyen, Duc Vu, Trung Dao, Chi Tran, Toan Tran, Anh Tran

Abstract: Shortcut models represent a promising, non-adversarial paradigm for generative modeling, uniquely supporting one-step, few-step, and multi-step sampling from a single trained network. However, their widespread adoption has been stymied by critical performance bottlenecks. This paper tackles the five core issues that held shortcut models back: (1) the hidden flaw of compounding guidance, which we a… ▽ More Shortcut models represent a promising, non-adversarial paradigm for generative modeling, uniquely supporting one-step, few-step, and multi-step sampling from a single trained network. However, their widespread adoption has been stymied by critical performance bottlenecks. This paper tackles the five core issues that held shortcut models back: (1) the hidden flaw of compounding guidance, which we are the first to formalize, causing severe image artifacts; (2) inflexible fixed guidance that restricts inference-time control; (3) a pervasive frequency bias driven by a reliance on low-level distances in the direct domain, which biases reconstructions toward low frequencies; (4) divergent self-consistency arising from a conflict with EMA training; and (5) curvy flow trajectories that impede convergence. To address these challenges, we introduce iSM, a unified training framework that systematically resolves each limitation. Our framework is built on four key improvements: Intrinsic Guidance provides explicit, dynamic control over guidance strength, resolving both compounding guidance and inflexibility. A Multi-Level Wavelet Loss mitigates frequency bias to restore high-frequency details. Scaling Optimal Transport (sOT) reduces training variance and learns straighter, more stable generative paths. Finally, a Twin EMA strategy reconciles training stability with self-consistency. Extensive experiments on ImageNet 256 x 256 demonstrate that our approach yields substantial FID improvements over baseline shortcut models across one-step, few-step, and multi-step generation, making shortcut models a viable and competitive class of generative models. △ Less

Submitted 24 October, 2025; originally announced October 2025.

Comments: Accepted at NeurIPS 2025

LubDubDecoder: Bringing Micro-Mechanical Cardiac Monitoring to Hearables

Authors: Siqi Zhang, Xiyuxing Zhang, Duc Vu, Tao Qiang, Clara Palacios, Jiangyifei Zhu, Yuntao Wang, Mayank Goel, Justin Chan

Abstract: We present LubDubDecoder, a system that enables fine-grained monitoring of micro-cardiac vibrations associated with the opening and closing of heart valves across a range of hearables. Our system transforms the built-in speaker, the only transducer common to all hearables, into an acoustic sensor that captures the coarse "lub-dub" heart sounds, leverages their shared temporal and spectral structur… ▽ More We present LubDubDecoder, a system that enables fine-grained monitoring of micro-cardiac vibrations associated with the opening and closing of heart valves across a range of hearables. Our system transforms the built-in speaker, the only transducer common to all hearables, into an acoustic sensor that captures the coarse "lub-dub" heart sounds, leverages their shared temporal and spectral structure to reconstruct the subtle seismocardiography (SCG) and gyrocardiography (GCG) waveforms, and extract the timing of key micro-cardiac events. In an IRB-approved feasibility study with 25 users, our system achieves correlations of 0.88-0.95 compared to chest-mounted reference measurements in within-user and cross-user evaluations, and generalizes to unseen hearables using a zero-effort adaptation scheme with a correlation of 0.91. Our system is robust across remounting sessions and music playback. △ Less

Submitted 8 April, 2026; v1 submitted 12 September, 2025; originally announced September 2025.

Fact-Checking at Scale: Multimodal AI for Authenticity and Context Verification in Online Media

Authors: Van-Hoang Phan, Tung-Duong Le-Duc, Long-Khanh Pham, Anh-Thu Le, Quynh-Huong Dinh-Nguyen, Dang-Quan Vo, Hoang-Quoc Nguyen-Son, Anh-Duy Tran, Dang Vu, Minh-Son Dao

Abstract: The proliferation of multimedia content on social media platforms has dramatically transformed how information is consumed and disseminated. While this shift enables real-time coverage of global events, it also facilitates the rapid spread of misinformation and disinformation, especially during crises such as wars, natural disasters, or elections. The rise of synthetic media and the reuse of authe… ▽ More The proliferation of multimedia content on social media platforms has dramatically transformed how information is consumed and disseminated. While this shift enables real-time coverage of global events, it also facilitates the rapid spread of misinformation and disinformation, especially during crises such as wars, natural disasters, or elections. The rise of synthetic media and the reuse of authentic content in misleading contexts have intensified the need for robust multimedia verification tools. In this paper, we present a comprehensive system developed for the ACM Multimedia 2025 Grand Challenge on Multimedia Verification. Our system assesses the authenticity and contextual accuracy of multimedia content in multilingual settings and generates both expert-oriented verification reports and accessible summaries for the general public. We introduce a unified verification pipeline that integrates visual forensics, textual analysis, and multimodal reasoning, and propose a hybrid approach to detect out-of-context (OOC) media through semantic similarity, temporal alignment, and geolocation cues. Extensive evaluations on the Grand Challenge benchmark demonstrate the system's effectiveness across diverse real-world scenarios. Our contributions advance the state of the art in multimedia verification and offer practical tools for journalists, fact-checkers, and researchers confronting information integrity challenges in the digital age. △ Less

Submitted 4 October, 2025; v1 submitted 11 August, 2025; originally announced August 2025.

Comments: Serious errors in results, will not be corrected

TestWeaver: Execution-aware, Feedback-driven Regression Testing Generation with Large Language Models

Authors: Cuong Chi Le, Cuong Duc Van, Tung Duy Vu, Thai Minh Pham Vu, Hoang Nhat Phan, Huy Nhat Phan, Tien N. Nguyen

Abstract: While recent advances in large language models (LLMs) have shown promise in automating test generation for regression testing, they often suffer from limited reasoning about program execution, resulting in stagnated coverage growth - a phenomenon known as the coverage plateau. This paper presents TestWeaver, a novel LLM-based approach that integrates lightweight program analysis to create a focuse… ▽ More While recent advances in large language models (LLMs) have shown promise in automating test generation for regression testing, they often suffer from limited reasoning about program execution, resulting in stagnated coverage growth - a phenomenon known as the coverage plateau. This paper presents TestWeaver, a novel LLM-based approach that integrates lightweight program analysis to create a focused execution context that assists LLMs in better test generation. TestWeaver strategically chooses the following components to overcome LLMs' limited reasoning on complex execution: (1) it reduces hallucinations and improves focus by supplying the LLM with the backward slice from the target line instead of a full program context; (2) it identifies and incorporates close test cases - those that share control-flow similarities with the path to the target line - to provide focused execution context within the LLM's context window; and (3) it enhances LLM's reasoning with execution in-line annotations that encode variable states as comments along the executed path. By equipping LLMs with these targeted and contextualized inputs, it improves coverage-guided test generation and mitigates redundant explorations. Empirical results show that TestWeaver accelerates code coverage growth and generates more effective test cases than the state-of-the-art approaches. △ Less

Submitted 26 January, 2026; v1 submitted 2 August, 2025; originally announced August 2025.

Comments: Accepted in ICSE 2026

Lucy: edgerunning agentic web search on mobile with machine generated task vectors

Authors: Alan Dao, Dinh Bach Vu, Alex Nguyen, Norapat Buppodom

Abstract: Small language models (SLMs) are inherently limited in knowledge-intensive tasks due to their constrained capacity. While test-time computation offers a path to enhanced performance, most approaches treat reasoning as a fixed or heuristic process. In this work, we propose a new paradigm: viewing the model's internal reasoning, delimited by <think> and </think> tags, as a dynamic task vector machin… ▽ More Small language models (SLMs) are inherently limited in knowledge-intensive tasks due to their constrained capacity. While test-time computation offers a path to enhanced performance, most approaches treat reasoning as a fixed or heuristic process. In this work, we propose a new paradigm: viewing the model's internal reasoning, delimited by <think> and </think> tags, as a dynamic task vector machine. Rather than treating the content inside these tags as a mere trace of thought, we interpret the generation process itself as a mechanism through which the model \textbf{constructs and refines its own task vectors} on the fly. We developed a method to optimize this dynamic task vector machine through RLVR and successfully trained an agentic web-search model. We present Lucy, a 1.7B-parameter SLM that leverages this dynamic reasoning mechanism with MCP integration to achieve 78.3% accuracy on the SimpleQA benchmark, performing on par with much larger models such as DeepSeek-V3. This demonstrates that small models can rival large ones when equipped with structured, self-constructed task reasoning. △ Less

Submitted 1 August, 2025; originally announced August 2025.

Queue up for takeoff: a transferable deep learning framework for flight delay prediction

Authors: Nnamdi Daniel Aghanya, Ta Duong Vu, Amaëlle Diop, Charlotte Deville, Nour Imane Kerroumi, Irene Moulitsas, Jun Li, Desmond Bisandu

Abstract: Flight delays are a significant challenge in the aviation industry, causing major financial and operational disruptions. To improve passenger experience and reduce revenue loss, flight delay prediction models must be both precise and generalizable across different networks. This paper introduces a novel approach that combines Queue-Theory with a simple attention model, referred to as the Queue-The… ▽ More Flight delays are a significant challenge in the aviation industry, causing major financial and operational disruptions. To improve passenger experience and reduce revenue loss, flight delay prediction models must be both precise and generalizable across different networks. This paper introduces a novel approach that combines Queue-Theory with a simple attention model, referred to as the Queue-Theory SimAM (QT-SimAM). To validate our model, we used data from the US Bureau of Transportation Statistics, where our proposed QT-SimAM (Bidirectional) model outperformed existing methods with an accuracy of 0.927 and an F1 score of 0.932. To assess transferability, we tested the model on the EUROCONTROL dataset. The results demonstrated strong performance, achieving an accuracy of 0.826 and an F1 score of 0.791. Ultimately, this paper outlines an effective, end-to-end methodology for predicting flight delays. The proposed model's ability to forecast delays with high accuracy across different networks can help reduce passenger anxiety and improve operational decision-making △ Less

Submitted 11 July, 2025; originally announced July 2025.

Comments: 3 figures, 20 pages references and appendix included,

MSC Class: 68T07; 90B22; 62M10 ACM Class: I.2.m

VergeIO: Depth-Aware Eye Interaction on Glasses

Authors: Xiyuxing Zhang, Duc Vu, Chengyi Shen, Yuntao Wang, Yuanchun Shi, Justin Chan

Abstract: There is growing industry interest in creating unobtrusive designs for electrooculography (EOG) sensing of eye gestures on glasses (e.g. JINS MEME and Apple eyewear). We present VergeIO, the first EOG-based glasses that enables depth-aware eye interaction using vergence with an optimized electrode layout and novel smart glass prototype. It can distinguish between four and six depth-based eye gestu… ▽ More There is growing industry interest in creating unobtrusive designs for electrooculography (EOG) sensing of eye gestures on glasses (e.g. JINS MEME and Apple eyewear). We present VergeIO, the first EOG-based glasses that enables depth-aware eye interaction using vergence with an optimized electrode layout and novel smart glass prototype. It can distinguish between four and six depth-based eye gestures with 83-98% accuracy using personalized models in a user study across 20 users and 2,400 gesture instances. It generalizes to unseen users with an accuracy of 77-97% without any calibration. To reduce false detections, we incorporate a motion artifact detection pipeline and a preamble-based activation scheme. The system uses dry sensors without any adhesives or gel, and operates in real time with 3 mW power consumption by the sensing front-end, making it suitable for always-on sensing. △ Less

Submitted 8 February, 2026; v1 submitted 2 July, 2025; originally announced July 2025.

Jan-nano Technical Report

Authors: Alan Dao, Dinh Bach Vu

Abstract: Most language models face a fundamental tradeoff where powerful capabilities require substantial computational resources. We shatter this constraint with Jan-nano, a 4B parameter language model that redefines efficiency through radical specialization: instead of trying to know everything, it masters the art of finding anything instantly. Fine-tuned from Qwen3-4B using our novel multi-stage Reinfor… ▽ More Most language models face a fundamental tradeoff where powerful capabilities require substantial computational resources. We shatter this constraint with Jan-nano, a 4B parameter language model that redefines efficiency through radical specialization: instead of trying to know everything, it masters the art of finding anything instantly. Fine-tuned from Qwen3-4B using our novel multi-stage Reinforcement Learning with Verifiable Rewards (RLVR) system that completely eliminates reliance on next token prediction training (SFT), Jan-nano achieves 83.2% on SimpleQA benchmark with MCP integration while running on consumer hardware. With 128K context length, Jan-nano proves that intelligence isn't about scale, it's about strategy. △ Less

Submitted 14 July, 2025; v1 submitted 28 June, 2025; originally announced June 2025.

E-FreeM2: Efficient Training-Free Multi-Scale and Cross-Modal News Verification via MLLMs

Authors: Van-Hoang Phan, Long-Khanh Pham, Dang Vu, Anh-Duy Tran, Minh-Son Dao

Abstract: The rapid spread of misinformation in mobile and wireless networks presents critical security challenges. This study introduces a training-free, retrieval-based multimodal fact verification system that leverages pretrained vision-language models and large language models for credibility assessment. By dynamically retrieving and cross-referencing trusted data sources, our approach mitigates vulnera… ▽ More The rapid spread of misinformation in mobile and wireless networks presents critical security challenges. This study introduces a training-free, retrieval-based multimodal fact verification system that leverages pretrained vision-language models and large language models for credibility assessment. By dynamically retrieving and cross-referencing trusted data sources, our approach mitigates vulnerabilities of traditional training-based models, such as adversarial attacks and data poisoning. Additionally, its lightweight design enables seamless edge device integration without extensive on-device processing. Experiments on two fact-checking benchmarks achieve SOTA results, confirming its effectiveness in misinformation detection and its robustness against various attack vectors, highlighting its potential to enhance security in mobile and wireless communication environments. △ Less

Submitted 25 June, 2025; originally announced June 2025.

Comments: Accepted to AsiaCCS 2025 @ SCID

MonoVQD: Monocular 3D Object Detection with Variational Query Denoising and Self-Distillation

Authors: Kiet Dang Vu, Trung Thai Tran, Duc Dung Nguyen

Abstract: Precisely localizing 3D objects from a single image constitutes a central challenge in monocular 3D detection. While DETR-like architectures offer a powerful paradigm, their direct application in this domain encounters inherent limitations, preventing optimal performance. Our work addresses these challenges by introducing MonoVQD, a novel framework designed to fundamentally advance DETR-based mono… ▽ More Precisely localizing 3D objects from a single image constitutes a central challenge in monocular 3D detection. While DETR-like architectures offer a powerful paradigm, their direct application in this domain encounters inherent limitations, preventing optimal performance. Our work addresses these challenges by introducing MonoVQD, a novel framework designed to fundamentally advance DETR-based monocular 3D detection. We propose three main contributions. First, we propose the Mask Separated Self-Attention mechanism that enables the integration of the denoising process into a DETR architecture. This improves the stability of Hungarian matching to achieve a consistent optimization objective. Second, we present the Variational Query Denoising technique to address the gradient vanishing problem of conventional denoising methods, which severely restricts the efficiency of the denoising process. This explicitly introduces stochastic properties to mitigate this fundamental limitation and unlock substantial performance gains. Finally, we introduce a sophisticated self-distillation strategy, leveraging insights from later decoder layers to synergistically improve query quality in earlier layers, thereby amplifying the iterative refinement process. Rigorous experimentation demonstrates that MonoVQD achieves superior performance on the challenging KITTI monocular benchmark. Highlighting its broad applicability, MonoVQD's core components seamlessly integrate into other architectures, delivering significant performance gains even in multi-view 3D detection scenarios on the nuScenes dataset and underscoring its robust generalization capabilities. △ Less

Submitted 14 June, 2025; originally announced June 2025.

Feeling Machines: Ethics, Culture, and the Rise of Emotional AI

Authors: Vivek Chavan, Arsen Cenaj, Shuyuan Shen, Ariane Bar, Srishti Binwani, Tommaso Del Becaro, Marius Funk, Lynn Greschner, Roberto Hung, Stina Klein, Romina Kleiner, Stefanie Krause, Sylwia Olbrych, Vishvapalsinhji Parmar, Jaleh Sarafraz, Daria Soroko, Daksitha Withanage Don, Chang Zhou, Hoang Thuy Duong Vu, Parastoo Semnani, Daniel Weinhardt, Elisabeth Andre, Jörg Krüger, Xavier Fresquet

Abstract: This paper explores the growing presence of emotionally responsive artificial intelligence through a critical and interdisciplinary lens. Bringing together the voices of early-career researchers from multiple fields, it explores how AI systems that simulate or interpret human emotions are reshaping our interactions in areas such as education, healthcare, mental health, caregiving, and digital life… ▽ More This paper explores the growing presence of emotionally responsive artificial intelligence through a critical and interdisciplinary lens. Bringing together the voices of early-career researchers from multiple fields, it explores how AI systems that simulate or interpret human emotions are reshaping our interactions in areas such as education, healthcare, mental health, caregiving, and digital life. The analysis is structured around four central themes: the ethical implications of emotional AI, the cultural dynamics of human-machine interaction, the risks and opportunities for vulnerable populations, and the emerging regulatory, design, and technical considerations. The authors highlight the potential of affective AI to support mental well-being, enhance learning, and reduce loneliness, as well as the risks of emotional manipulation, over-reliance, misrepresentation, and cultural bias. Key challenges include simulating empathy without genuine understanding, encoding dominant sociocultural norms into AI systems, and insufficient safeguards for individuals in sensitive or high-risk contexts. Special attention is given to children, elderly users, and individuals with mental health challenges, who may interact with AI in emotionally significant ways. However, there remains a lack of cognitive or legal protections which are necessary to navigate such engagements safely. The report concludes with ten recommendations, including the need for transparency, certification frameworks, region-specific fine-tuning, human oversight, and longitudinal research. A curated supplementary section provides practical tools, models, and datasets to support further work in this domain. △ Less

Submitted 14 June, 2025; originally announced June 2025.

Comments: From the Spring School 2025 by AI Grid and SCAI (Sorbonne University), 16 pages

The RSNA Lumbar Degenerative Imaging Spine Classification (LumbarDISC) Dataset

Authors: Tyler J. Richards, Adam E. Flanders, Errol Colak, Luciano M. Prevedello, Robyn L. Ball, Felipe Kitamura, John Mongan, Maryam Vazirabad, Hui-Ming Lin, Anne Kendell, Thanat Kanthawang, Salita Angkurawaranon, Emre Altinmakas, Hakan Dogan, Paulo Eduardo de Aguiar Kuriki, Arjuna Somasundaram, Christopher Ruston, Deniz Bulja, Naida Spahovic, Jennifer Sommer, Sirui Jiang, Eduardo Moreno Judice de Mattos Farina, Eduardo Caminha Nunes, Michael Brassil, Megan McNamara , et al. (11 additional authors not shown)

Abstract: The Radiological Society of North America (RSNA) Lumbar Degenerative Imaging Spine Classification (LumbarDISC) dataset is the largest publicly available dataset of adult MRI lumbar spine examinations annotated for degenerative changes. The dataset includes 2,697 patients with a total of 8,593 image series from 8 institutions across 6 countries and 5 continents. The dataset is available for free fo… ▽ More The Radiological Society of North America (RSNA) Lumbar Degenerative Imaging Spine Classification (LumbarDISC) dataset is the largest publicly available dataset of adult MRI lumbar spine examinations annotated for degenerative changes. The dataset includes 2,697 patients with a total of 8,593 image series from 8 institutions across 6 countries and 5 continents. The dataset is available for free for non-commercial use via Kaggle and RSNA Medical Imaging Resource of AI (MIRA). The dataset was created for the RSNA 2024 Lumbar Spine Degenerative Classification competition where competitors developed deep learning models to grade degenerative changes in the lumbar spine. The degree of spinal canal, subarticular recess, and neural foraminal stenosis was graded at each intervertebral disc level in the lumbar spine. The images were annotated by expert volunteer neuroradiologists and musculoskeletal radiologists from the RSNA, American Society of Neuroradiology, and the American Society of Spine Radiology. This dataset aims to facilitate research and development in machine learning and lumbar spine imaging to lead to improved patient care and clinical efficiency. △ Less

Submitted 10 June, 2025; originally announced June 2025.

Autoencoding Random Forests

Authors: Binh Duc Vu, Jan Kapar, Marvin Wright, David S. Watson

Abstract: We propose a principled method for autoencoding with random forests. Our strategy builds on foundational results from nonparametric statistics and spectral graph theory to learn a low-dimensional embedding of the model that optimally represents relationships in the data. We provide exact and approximate solutions to the decoding problem via constrained optimization, split relabeling, and nearest n… ▽ More We propose a principled method for autoencoding with random forests. Our strategy builds on foundational results from nonparametric statistics and spectral graph theory to learn a low-dimensional embedding of the model that optimally represents relationships in the data. We provide exact and approximate solutions to the decoding problem via constrained optimization, split relabeling, and nearest neighbors regression. These methods effectively invert the compression pipeline, establishing a map from the embedding space back to the input space using splits learned by the ensemble's constituent trees. The resulting decoders are universally consistent under common regularity assumptions. The procedure works with supervised or unsupervised models, providing a window into conditional or joint distributions. We demonstrate various applications of this autoencoder, including powerful new tools for visualization, compression, clustering, and denoising. Experiments illustrate the ease and utility of our method in a wide range of settings, including tabular, image, and genomic data. △ Less

Submitted 14 January, 2026; v1 submitted 27 May, 2025; originally announced May 2025.

Comments: 10 pages main text, 27 pages total. 9 figures, 4 tables. To be published in proceedings of the 39th Conference on Neural Information Processing Systems (NeurIPS 2025)

Speechless: Speech Instruction Training Without Speech for Low Resource Languages

Authors: Alan Dao, Dinh Bach Vu, Huy Hoang Ha, Tuan Le Duc Anh, Shreyas Gopal, Yue Heng Yeo, Warren Keng Hoong Low, Eng Siong Chng, Jia Qi Yip

Abstract: The rapid growth of voice assistants powered by large language models (LLM) has highlighted a need for speech instruction data to train these systems. Despite the abundance of speech recognition data, there is a notable scarcity of speech instruction data, which is essential for fine-tuning models to understand and execute spoken commands. Generating high-quality synthetic speech requires a good t… ▽ More The rapid growth of voice assistants powered by large language models (LLM) has highlighted a need for speech instruction data to train these systems. Despite the abundance of speech recognition data, there is a notable scarcity of speech instruction data, which is essential for fine-tuning models to understand and execute spoken commands. Generating high-quality synthetic speech requires a good text-to-speech (TTS) model, which may not be available to low resource languages. Our novel approach addresses this challenge by halting synthesis at the semantic representation level, bypassing the need for TTS. We achieve this by aligning synthetic semantic representations with the pre-trained Whisper encoder, enabling an LLM to be fine-tuned on text instructions while maintaining the ability to understand spoken instructions during inference. This simplified training process is a promising approach to building voice assistant for low-resource languages. △ Less

Submitted 22 May, 2025; originally announced May 2025.

Comments: This paper was accepted by INTERSPEECH 2025

Journal ref: Proc. Interspeech 2025, 3239-3243

Out-of-Distribution Detection via Channelwise Feature Aggregation in Neural Network-Based Receivers

Authors: Marko Tuononen, Heikki Penttinen, Duy Vu, Dani Korpi, Vesa Starck, Ville Hautamäki

Abstract: Neural network-based radio receivers are expected to play a key role in future wireless systems, making reliable Out-Of-Distribution (OOD) detection essential. We propose a post-hoc, layerwise OOD framework based on channelwise feature aggregation that avoids classwise statistics--critical for multi-label soft-bit outputs with astronomically many classes. Receiver activations exhibit no discrete c… ▽ More Neural network-based radio receivers are expected to play a key role in future wireless systems, making reliable Out-Of-Distribution (OOD) detection essential. We propose a post-hoc, layerwise OOD framework based on channelwise feature aggregation that avoids classwise statistics--critical for multi-label soft-bit outputs with astronomically many classes. Receiver activations exhibit no discrete clusters but a smooth Signal-to-Noise-Ratio (SNR)-aligned manifold, consistent with classical receiver behavior and motivating manifold-aware OOD detection. We evaluate multiple OOD feature types, distance metrics, and methods across layers. Gaussian Mahalanobis with mean activations is the strongest single detector, earlier layers outperform later, and SNR/classifier fusions offer small, inconsistent AUROC gains. High-delay OOD is detected reliably, while high-speed remains challenging. △ Less

Submitted 14 January, 2026; v1 submitted 21 May, 2025; originally announced May 2025.

Comments: 50 pages, 39 figures, 48 tables, and 31 equations

MSC Class: 68T07 (Primary) 62H30; 62G05; 94A05; 68T30 (Secondary) ACM Class: I.2.6; I.5.1; C.2.3

Multimodal graph representation learning for website generation based on visual sketch

Authors: Tung D. Vu, Chung Hoang, Truong-Son Hy

Abstract: The Design2Code problem, which involves converting digital designs into functional source code, is a significant challenge in software development due to its complexity and time-consuming nature. Traditional approaches often struggle with accurately interpreting the intricate visual details and structural relationships inherent in webpage designs, leading to limitations in automation and efficienc… ▽ More The Design2Code problem, which involves converting digital designs into functional source code, is a significant challenge in software development due to its complexity and time-consuming nature. Traditional approaches often struggle with accurately interpreting the intricate visual details and structural relationships inherent in webpage designs, leading to limitations in automation and efficiency. In this paper, we propose a novel method that leverages multimodal graph representation learning to address these challenges. By integrating both visual and structural information from design sketches, our approach enhances the accuracy and efficiency of code generation, particularly in producing semantically correct and structurally sound HTML code. We present a comprehensive evaluation of our method, demonstrating significant improvements in both accuracy and efficiency compared to existing techniques. Extensive evaluation demonstrates significant improvements of multimodal graph learning over existing techniques, highlighting the potential of our method to revolutionize design-to-code automation. Code available at https://github.com/HySonLab/Design2Code △ Less

Submitted 25 April, 2025; originally announced April 2025.

SuoiAI: Building a Dataset for Aquatic Invertebrates in Vietnam

Authors: Tue Vo, Lakshay Sharma, Tuan Dinh, Khuong Dinh, Trang Nguyen, Trung Phan, Minh Do, Duong Vu

Abstract: Understanding and monitoring aquatic biodiversity is critical for ecological health and conservation efforts. This paper proposes SuoiAI, an end-to-end pipeline for building a dataset of aquatic invertebrates in Vietnam and employing machine learning (ML) techniques for species classification. We outline the methods for data collection, annotation, and model training, focusing on reducing annotati… ▽ More Understanding and monitoring aquatic biodiversity is critical for ecological health and conservation efforts. This paper proposes SuoiAI, an end-to-end pipeline for building a dataset of aquatic invertebrates in Vietnam and employing machine learning (ML) techniques for species classification. We outline the methods for data collection, annotation, and model training, focusing on reducing annotation effort through semi-supervised learning and leveraging state-of-the-art object detection and classification models. Our approach aims to overcome challenges such as data scarcity, fine-grained classification, and deployment in diverse environmental conditions. △ Less

Submitted 21 April, 2025; originally announced April 2025.

Comments: Published as a workshop paper at "Tackling Climate Change with Machine Learning", ICLR 2025

FaR: Enhancing Multi-Concept Text-to-Image Diffusion via Concept Fusion and Localized Refinement

Authors: Gia-Nghia Tran, Quang-Huy Che, Trong-Tai Dam Vu, Bich-Nga Pham, Vinh-Tiep Nguyen, Trung-Nghia Le, Minh-Triet Tran

Abstract: Generating multiple new concepts remains a challenging problem in the text-to-image task. Current methods often overfit when trained on a small number of samples and struggle with attribute leakage, particularly for class-similar subjects (e.g., two specific dogs). In this paper, we introduce Fuse-and-Refine (FaR), a novel approach that tackles these challenges through two key contributions: Conce… ▽ More Generating multiple new concepts remains a challenging problem in the text-to-image task. Current methods often overfit when trained on a small number of samples and struggle with attribute leakage, particularly for class-similar subjects (e.g., two specific dogs). In this paper, we introduce Fuse-and-Refine (FaR), a novel approach that tackles these challenges through two key contributions: Concept Fusion technique and Localized Refinement loss function. Concept Fusion systematically augments the training data by separating reference subjects from backgrounds and recombining them into composite images to increase diversity. This augmentation technique tackles the overfitting problem by mitigating the narrow distribution of the limited training samples. In addition, Localized Refinement loss function is introduced to preserve subject representative attributes by aligning each concept's attention map to its correct region. This approach effectively prevents attribute leakage by ensuring that the diffusion model distinguishes similar subjects without mixing their attention maps during the denoising process. By fine-tuning specific modules at the same time, FaR balances the learning of new concepts with the retention of previously learned knowledge. Empirical results show that FaR not only prevents overfitting and attribute leakage while maintaining photorealism, but also outperforms other state-of-the-art methods. △ Less

Submitted 4 April, 2025; originally announced April 2025.

VNJPTranslate: A comprehensive pipeline for Vietnamese-Japanese translation

Authors: Hoang Hai Phan, Nguyen Duc Minh Vu, Nam Dang Phuong

Abstract: Neural Machine Translation (NMT) driven by Transformer architectures has advanced significantly, yet faces challenges with low-resource language pairs like Vietnamese-Japanese (Vi-Ja). Issues include sparse parallel data and handling linguistic/cultural nuances. Recent progress in Large Language Models (LLMs) with strong reasoning, often refined via Reinforcement Learning (RL), enables high-qualit… ▽ More Neural Machine Translation (NMT) driven by Transformer architectures has advanced significantly, yet faces challenges with low-resource language pairs like Vietnamese-Japanese (Vi-Ja). Issues include sparse parallel data and handling linguistic/cultural nuances. Recent progress in Large Language Models (LLMs) with strong reasoning, often refined via Reinforcement Learning (RL), enables high-quality synthetic data generation. We introduce VNJPTranslate, a pipeline designed to systematically address the Vi-Ja translation task. It features a targeted data augmentation strategy using advanced LLMs with Chain-of-Thought prompting for challenging segments identified via corpus analysis. Subsequently, we employ efficient fine-tuning techniques (Unsloth with QLoRA) on a capable, low-parameter autoregressive model (specifically, a fine-tuned version of the 1.8B parameter Sailor model, which is based on the Qwen architecture) to create a practical and high-performing translation system. This integrated approach aims to improve Vi-Ja translation quality significantly over existing baselines. △ Less

Submitted 12 October, 2025; v1 submitted 31 March, 2025; originally announced April 2025.

Comments: The paper contains a critical error in Section 3.1, leading to invalid results in Section 3.3. This undermines the main conclusion of the paper. The authors are working on a corrected version, but in the meantime, there is not a quick fix/replacement/update available

AlphaSpace: Enabling Robotic Actions through Semantic Tokenization and Symbolic Reasoning

Authors: Alan Dao, Dinh Bach Vu, Bui Quang Huy

Abstract: This paper presents AlphaSpace, a novel methodology designed to enhance the spatial reasoning capabilities of language models for robotic manipulation in 3D Cartesian space. AlphaSpace employs a hierarchical semantics-based tokenization strategy that encodes spatial information at both coarse and fine-grained levels. Our approach represents objects with their attributes, positions, and height info… ▽ More This paper presents AlphaSpace, a novel methodology designed to enhance the spatial reasoning capabilities of language models for robotic manipulation in 3D Cartesian space. AlphaSpace employs a hierarchical semantics-based tokenization strategy that encodes spatial information at both coarse and fine-grained levels. Our approach represents objects with their attributes, positions, and height information through structured tokens, enabling precise spatial reasoning without relying on traditional vision-based embeddings. This approach enables LLMs to accurately manipulate objects by positioning them at specific (x, y, z) coordinates. Experimental results suggest that AlphaSpace demonstrates promising potential for improving manipulation tasks, achieving a total accuracy of 66.67%, compared to 37.5% for GPT-4o and 29.17% for Claude 3.5 Sonnet. These results demonstrate the potential of structured spatial encoding for manipulation tasks and warrant further exploration. △ Less

Submitted 27 March, 2025; v1 submitted 24 March, 2025; originally announced March 2025.

Explainable Graph-theoretical Machine Learning: with Application to Alzheimer's Disease Prediction

Authors: Narmina Baghirova, Duy-Thanh Vũ, Duy-Cat Can, Christelle Schneuwly Diaz, Julien Bodlet, Guillaume Blanc, Georgi Hrusanov, Bernard Ries, Oliver Y. Chén

Abstract: Alzheimer's disease (AD) affects 50 million people worldwide and is projected to overwhelm 152 million by 2050. AD is characterized by cognitive decline due partly to disruptions in metabolic brain connectivity. Thus, early and accurate detection of metabolic brain network impairments is crucial for AD management. Chief to identifying such impairments is FDG-PET data. Despite advancements, most gr… ▽ More Alzheimer's disease (AD) affects 50 million people worldwide and is projected to overwhelm 152 million by 2050. AD is characterized by cognitive decline due partly to disruptions in metabolic brain connectivity. Thus, early and accurate detection of metabolic brain network impairments is crucial for AD management. Chief to identifying such impairments is FDG-PET data. Despite advancements, most graph-based studies using FDG-PET data rely on group-level analysis or thresholding. Yet, group-level analysis can veil individual differences and thresholding may overlook weaker but biologically critical brain connections. Additionally, machine learning-based AD prediction largely focuses on univariate outcomes, such as disease status. Here, we introduce explainable graph-theoretical machine learning (XGML), a framework employing kernel density estimation and dynamic time warping to construct individual metabolic brain graphs that capture the distance between pair-wise brain regions and identify subgraphs most predictive of multivariate AD-related outcomes. Using FDG-PET data from the Alzheimer's Disease Neuroimaging Initiative, XGML builds metabolic brain graphs and uncovers subgraphs predictive of eight AD-related cognitive scores in new subjects. XGML shows robust performance, particularly for predicting scores measuring learning, memory, language, praxis, and orientation, such as CDRSB ($r = 0.74$), ADAS11 ($r = 0.73$), and ADAS13 ($r = 0.71$). Moreover, XGML unveils key edges jointly but differentially predictive of several AD-related outcomes; they may serve as potential network biomarkers for assessing overall cognitive decline. Together, we show the promise of graph-theoretical machine learning in biomarker discovery and disease prediction and its potential to improve our understanding of network neural mechanisms underlying AD. △ Less

Submitted 20 March, 2025; originally announced March 2025.

OPTIMUS: Predicting Multivariate Outcomes in Alzheimer's Disease Using Multi-modal Data amidst Missing Values

Authors: Christelle Schneuwly Diaz, Duy-Thanh Vu, Julien Bodelet, Duy-Cat Can, Guillaume Blanc, Haiting Jiang, Lin Yao, Guiseppe Pantaleo, ADNI, Oliver Y. Chén

Abstract: Alzheimer's disease, a neurodegenerative disorder, is associated with neural, genetic, and proteomic factors while affecting multiple cognitive and behavioral faculties. Traditional AD prediction largely focuses on univariate disease outcomes, such as disease stages and severity. Multimodal data encode broader disease information than a single modality and may, therefore, improve disease predictio… ▽ More Alzheimer's disease, a neurodegenerative disorder, is associated with neural, genetic, and proteomic factors while affecting multiple cognitive and behavioral faculties. Traditional AD prediction largely focuses on univariate disease outcomes, such as disease stages and severity. Multimodal data encode broader disease information than a single modality and may, therefore, improve disease prediction; but they often contain missing values. Recent "deeper" machine learning approaches show promise in improving prediction accuracy, yet the biological relevance of these models needs to be further charted. Integrating missing data analysis, predictive modeling, multimodal data analysis, and explainable AI, we propose OPTIMUS, a predictive, modular, and explainable machine learning framework, to unveil the many-to-many predictive pathways between multimodal input data and multivariate disease outcomes amidst missing values. OPTIMUS first applies modality-specific imputation to uncover data from each modality while optimizing overall prediction accuracy. It then maps multimodal biomarkers to multivariate outcomes using machine-learning and extracts biomarkers respectively predictive of each outcome. Finally, OPTIMUS incorporates XAI to explain the identified multimodal biomarkers. Using data from 346 cognitively normal subjects, 608 persons with mild cognitive impairment, and 251 AD patients, OPTIMUS identifies neural and transcriptomic signatures that jointly but differentially predict multivariate outcomes related to executive function, language, memory, and visuospatial function. Our work demonstrates the potential of building a predictive and biologically explainable machine-learning framework to uncover multimodal biomarkers that capture disease profiles across varying cognitive landscapes. The results improve our understanding of the complex many-to-many pathways in AD. △ Less

Submitted 14 March, 2025; originally announced March 2025.

PoseLess: Depth-Free Vision-to-Joint Control via Direct Image Mapping with VLM

Authors: Alan Dao, Dinh Bach Vu, Tuan Le Duc Anh, Bui Quang Huy

Abstract: This paper introduces PoseLess, a novel framework for robot hand control that eliminates the need for explicit pose estimation by directly mapping 2D images to joint angles using projected representations. Our approach leverages synthetic training data generated through randomized joint configurations, enabling zero-shot generalization to real-world scenarios and cross-morphology transfer from rob… ▽ More This paper introduces PoseLess, a novel framework for robot hand control that eliminates the need for explicit pose estimation by directly mapping 2D images to joint angles using projected representations. Our approach leverages synthetic training data generated through randomized joint configurations, enabling zero-shot generalization to real-world scenarios and cross-morphology transfer from robotic to human hands. By projecting visual inputs and employing a transformer-based decoder, PoseLess achieves robust, low-latency control while addressing challenges such as depth ambiguity and data scarcity. Experimental results demonstrate competitive performance in joint angle prediction accuracy without relying on any human-labelled dataset. △ Less

Submitted 10 March, 2025; v1 submitted 10 March, 2025; originally announced March 2025.

SuperRAG: Beyond RAG with Layout-Aware Graph Modeling

Authors: Jeff Yang, Duy-Khanh Vu, Minh-Tien Nguyen, Xuan-Quang Nguyen, Linh Nguyen, Hung Le

Abstract: This paper introduces layout-aware graph modeling for multimodal RAG. Different from traditional RAG methods that mostly deal with flat text chunks, the proposed method takes into account the relationship of multimodalities by using a graph structure. To do that, a graph modeling structure is defined based on document layout parsing. The structure of an input document is retained with the connecti… ▽ More This paper introduces layout-aware graph modeling for multimodal RAG. Different from traditional RAG methods that mostly deal with flat text chunks, the proposed method takes into account the relationship of multimodalities by using a graph structure. To do that, a graph modeling structure is defined based on document layout parsing. The structure of an input document is retained with the connection of text chunks, tables, and figures. This representation allows the method to handle complex questions that require information from multimodalities. To confirm the efficiency of the graph modeling, a flexible RAG pipeline is developed using robust components. Experimental results on four benchmark test sets confirm the contribution of the layout-aware modeling for performance improvement of the RAG pipeline. △ Less

Submitted 28 February, 2025; originally announced March 2025.

Comments: NAACL 2025, Industry Track

FactFlow: Automatic Fact Sheet Generation and Customization from Tabular Dataset via AI Chain Design & Implementation

Authors: Minh Duc Vu, Jieshan Chen, Zhenchang Xing, Qinghua Lu, Xiwei Xu, Qian Fu

Abstract: With the proliferation of data across various domains, there is a critical demand for tools that enable non-experts to derive meaningful insights without deep data analysis skills. To address this need, existing automatic fact sheet generation tools offer heuristic-based solutions to extract facts and generate stories. However, they inadequately grasp the semantics of data and struggle to generate… ▽ More With the proliferation of data across various domains, there is a critical demand for tools that enable non-experts to derive meaningful insights without deep data analysis skills. To address this need, existing automatic fact sheet generation tools offer heuristic-based solutions to extract facts and generate stories. However, they inadequately grasp the semantics of data and struggle to generate narratives that fully capture the semantics of the dataset or align the fact sheet with specific user needs. Addressing these shortcomings, this paper introduces \tool, a novel tool designed for the automatic generation and customisation of fact sheets. \tool applies the concept of collaborative AI workers to transform raw tabular dataset into comprehensive, visually compelling fact sheets. We define effective taxonomy to profile AI worker for specialised tasks. Furthermore, \tool empowers users to refine these fact sheets through intuitive natural language commands, ensuring the final outputs align closely with individual preferences and requirements. Our user evaluation with 18 participants confirms that \tool not only surpasses state-of-the-art baselines in automated fact sheet production but also provides a positive user experience during customization tasks. △ Less

Submitted 25 February, 2025; originally announced February 2025.

Comments: 11 pages, 6 figures

ACM Class: I.2; H.4

SQLong: Enhanced NL2SQL for Longer Contexts with LLMs

Authors: Dai Quoc Nguyen, Cong Duy Vu Hoang, Duy Vu, Gioacchino Tangari, Thanh Tien Vu, Don Dharmasiri, Yuan-Fang Li, Long Duong

Abstract: Open-weight large language models (LLMs) have significantly advanced performance in the Natural Language to SQL (NL2SQL) task. However, their effectiveness diminishes when dealing with large database schemas, as the context length increases. To address this limitation, we present SQLong, a novel and efficient data augmentation framework designed to enhance LLM performance in long-context scenarios… ▽ More Open-weight large language models (LLMs) have significantly advanced performance in the Natural Language to SQL (NL2SQL) task. However, their effectiveness diminishes when dealing with large database schemas, as the context length increases. To address this limitation, we present SQLong, a novel and efficient data augmentation framework designed to enhance LLM performance in long-context scenarios for the NL2SQL task. SQLong generates augmented datasets by extending existing database schemas with additional synthetic CREATE TABLE commands and corresponding data rows, sampled from diverse schemas in the training data. This approach effectively simulates long-context scenarios during finetuning and evaluation. Through experiments on the Spider and BIRD datasets, we demonstrate that LLMs finetuned with SQLong-augmented data significantly outperform those trained on standard datasets. These imply SQLong's practical implementation and its impact on improving NL2SQL capabilities in real-world settings with complex database schemas. △ Less

Submitted 20 May, 2025; v1 submitted 23 February, 2025; originally announced February 2025.

Comments: Accepted to Table Representation Learning Workshop at ACL 2025

AlphaMaze: Enhancing Large Language Models' Spatial Intelligence via GRPO

Authors: Alan Dao, Dinh Bach Vu

Abstract: Large Language Models (LLMs) have demonstrated impressive capabilities in language processing, yet they often struggle with tasks requiring genuine visual spatial reasoning. In this paper, we introduce a novel two-stage training framework designed to equip standard LLMs with visual reasoning abilities for maze navigation. First, we leverage Supervised Fine Tuning (SFT) on a curated dataset of toke… ▽ More Large Language Models (LLMs) have demonstrated impressive capabilities in language processing, yet they often struggle with tasks requiring genuine visual spatial reasoning. In this paper, we introduce a novel two-stage training framework designed to equip standard LLMs with visual reasoning abilities for maze navigation. First, we leverage Supervised Fine Tuning (SFT) on a curated dataset of tokenized maze representations to teach the model to predict step-by-step movement commands. Next, we apply Group Relative Policy Optimization (GRPO)-a technique used in DeepSeekR1-with a carefully crafted reward function to refine the model's sequential decision-making and encourage emergent chain-of-thought behaviors. Experimental results on synthetically generated mazes show that while a baseline model fails to navigate the maze, the SFT-trained model achieves 86% accuracy, and further GRPO fine-tuning boosts accuracy to 93%. Qualitative analyses reveal that GRPO fosters more robust and self-corrective reasoning, highlighting the potential of our approach to bridge the gap between language models and visual spatial tasks. These findings offer promising implications for applications in robotics, autonomous navigation, and other domains that require integrated visual and sequential reasoning. △ Less

Submitted 25 February, 2025; v1 submitted 20 February, 2025; originally announced February 2025.

CutPaste&Find: Efficient Multimodal Hallucination Detector with Visual-aid Knowledge Base

Authors: Cong-Duy Nguyen, Xiaobao Wu, Duc Anh Vu, Shuai Zhao, Thong Nguyen, Anh Tuan Luu

Abstract: Large Vision-Language Models (LVLMs) have demonstrated impressive multimodal reasoning capabilities, but they remain susceptible to hallucination, particularly object hallucination where non-existent objects or incorrect attributes are fabricated in generated descriptions. Existing detection methods achieve strong performance but rely heavily on expensive API calls and iterative LVLM-based validat… ▽ More Large Vision-Language Models (LVLMs) have demonstrated impressive multimodal reasoning capabilities, but they remain susceptible to hallucination, particularly object hallucination where non-existent objects or incorrect attributes are fabricated in generated descriptions. Existing detection methods achieve strong performance but rely heavily on expensive API calls and iterative LVLM-based validation, making them impractical for large-scale or offline use. To address these limitations, we propose CutPaste\&Find, a lightweight and training-free framework for detecting hallucinations in LVLM-generated outputs. Our approach leverages off-the-shelf visual and linguistic modules to perform multi-step verification efficiently without requiring LVLM inference. At the core of our framework is a Visual-aid Knowledge Base that encodes rich entity-attribute relationships and associated image representations. We introduce a scaling factor to refine similarity scores, mitigating the issue of suboptimal alignment values even for ground-truth image-text pairs. Comprehensive evaluations on benchmark datasets, including POPE and R-Bench, demonstrate that CutPaste\&Find achieves competitive hallucination detection performance while being significantly more efficient and cost-effective than previous methods. △ Less

Submitted 4 August, 2025; v1 submitted 18 February, 2025; originally announced February 2025.

A4O: All Trigger for One sample

Authors: Duc Anh Vu, Anh Tuan Tran, Cong Tran, Cuong Pham

Abstract: Backdoor attacks have become a critical threat to deep neural networks (DNNs), drawing many research interests. However, most of the studied attacks employ a single type of trigger. Consequently, proposed backdoor defenders often rely on the assumption that triggers would appear in a unified way. In this paper, we show that this naive assumption can create a loophole, allowing more sophisticated b… ▽ More Backdoor attacks have become a critical threat to deep neural networks (DNNs), drawing many research interests. However, most of the studied attacks employ a single type of trigger. Consequently, proposed backdoor defenders often rely on the assumption that triggers would appear in a unified way. In this paper, we show that this naive assumption can create a loophole, allowing more sophisticated backdoor attacks to bypass. We design a novel backdoor attack mechanism that incorporates multiple types of backdoor triggers, focusing on stealthiness and effectiveness. Our journey begins with the intriguing observation that the performance of a backdoor attack in deep learning models, as well as its detectability and removability, are all proportional to the magnitude of the trigger. Based on this correlation, we propose reducing the magnitude of each trigger type and combining them to achieve a strong backdoor relying on the combined trigger while still staying safely under the radar of defenders. Extensive experiments on three standard datasets demonstrate that our method can achieve high attack success rates (ASRs) while consistently bypassing state-of-the-art defenses. △ Less

Submitted 13 January, 2025; originally announced January 2025.

Curriculum Demonstration Selection for In-Context Learning

Authors: Duc Anh Vu, Nguyen Tran Cong Duy, Xiaobao Wu, Hoang Minh Nhat, Du Mingzhe, Nguyen Thanh Thong, Anh Tuan Luu

Abstract: Large Language Models (LLMs) have shown strong in-context learning (ICL) abilities with a few demonstrations. However, one critical challenge is how to select demonstrations to elicit the full potential of LLMs. In this paper, we propose Curriculum Demonstration Selection (CDS), a novel demonstration selection method for ICL. Instead of merely using similarity, CDS additionally partitions samples… ▽ More Large Language Models (LLMs) have shown strong in-context learning (ICL) abilities with a few demonstrations. However, one critical challenge is how to select demonstrations to elicit the full potential of LLMs. In this paper, we propose Curriculum Demonstration Selection (CDS), a novel demonstration selection method for ICL. Instead of merely using similarity, CDS additionally partitions samples by their complexity measurements. Following curriculum learning, CDS then selects demonstrations from easy to difficult. Thus the selected demonstrations cover a wide range of difficulty levels, enabling LLMs to learn from varied complexities within the training set. Experiments demonstrate that our CDS consistently outperforms baseline methods, achieving notable improvements across nine LLMs on three benchmarks. Moreover, CDS proves especially effective in enhancing LLM performance in solving challenging problems. △ Less

Submitted 15 December, 2024; v1 submitted 27 November, 2024; originally announced November 2024.

Comments: Accepted at the 40th ACM/SIGAPP Symposium On Applied Computing (SAC 2025), Main Conference

A Study of Malware Prevention in Linux Distributions

Authors: Duc-Ly Vu, Trevor Dunlap, Karla Obermeier-Velazquez, Thanh-Cong Nguyen, Paul Gibert, John Speed Meyers, Santiago Torres-Arias

Abstract: Malicious attacks on open-source software packages are a growing concern. The discovery of the XZ Utils backdoor intensified these concerns because of the potential widespread impact. This study, therefore, explores the challenges of preventing and detecting malware in Linux distribution package repositories. To do so, we ask two research questions: (1) What measures have Linux distributions imple… ▽ More Malicious attacks on open-source software packages are a growing concern. The discovery of the XZ Utils backdoor intensified these concerns because of the potential widespread impact. This study, therefore, explores the challenges of preventing and detecting malware in Linux distribution package repositories. To do so, we ask two research questions: (1) What measures have Linux distributions implemented to counter malware, and how have maintainers experienced these efforts? (2) How effective are current malware detection tools in identifying malicious Linux packages? To answer these questions, we conduct interviews with maintainers at several major Linux distributions and introduce a Linux package malware benchmark dataset. Using this dataset, we evaluate the performance of six open-source malware detection scanners. Distribution maintainers, according to the interviews, have mostly focused on reproducible builds to date. Our interviews identified only a single Linux distribution, Wolfi OS, that performs active malware scanning. Using this new benchmark dataset, the evaluation found that the performance of existing open-source malware scanners is underwhelming. Most studied tools excel at producing false positives but only infrequently detect true malware. Those that avoid high false positive rates often do so at the expense of a satisfactory true positive. Our findings provide insights into Linux distribution package repositories' current practices for malware detection and demonstrate the current inadequacy of open-source tools designed to detect malicious Linux packages. △ Less

Submitted 21 July, 2025; v1 submitted 17 November, 2024; originally announced November 2024.

Comments: 14 pages, 3 figures, 11 tables

Mastering the Craft of Data Synthesis for CodeLLMs

Authors: Meng Chen, Philip Arthur, Qianyu Feng, Cong Duy Vu Hoang, Yu-Heng Hong, Mahdi Kazemi Moghaddam, Omid Nezami, Thien Nguyen, Gioacchino Tangari, Duy Vu, Thanh Vu, Mark Johnson, Krishnaram Kenthapadi, Don Dharmasiri, Long Duong, Yuan-Fang Li

Abstract: Large language models (LLMs) have shown impressive performance in \emph{code} understanding and generation, making coding tasks a key focus for researchers due to their practical applications and value as a testbed for LLM evaluation. Data synthesis and filtering techniques have been widely adopted and shown to be highly effective in this context. In this paper, we present a focused survey and tax… ▽ More Large language models (LLMs) have shown impressive performance in \emph{code} understanding and generation, making coding tasks a key focus for researchers due to their practical applications and value as a testbed for LLM evaluation. Data synthesis and filtering techniques have been widely adopted and shown to be highly effective in this context. In this paper, we present a focused survey and taxonomy of these techniques, emphasizing recent advancements. We highlight key challenges, explore future research directions, and offer practical guidance for new researchers entering the field. △ Less

Submitted 7 February, 2025; v1 submitted 16 October, 2024; originally announced November 2024.

Comments: Accepted at NAACL 2025

Learning Graph Filters for Structure-Function Coupling based Hub Node Identification

Authors: Meiby Ortiz-Bouza, Duc Vu, Abdullah Karaaslanli, Selin Aviyente

Abstract: Over the past two decades, tools from network science have been leveraged to characterize the organization of both structural and functional networks of the brain. One such measure of network organization is hub node identification. Hubs are specialized nodes within a network that link distinct brain units corresponding to specialized functional processes. Conventional methods for identifying hub… ▽ More Over the past two decades, tools from network science have been leveraged to characterize the organization of both structural and functional networks of the brain. One such measure of network organization is hub node identification. Hubs are specialized nodes within a network that link distinct brain units corresponding to specialized functional processes. Conventional methods for identifying hub nodes utilize different types of centrality measures and participation coefficient to profile various aspects of nodal importance. These methods solely rely on the functional connectivity networks constructed from functional magnetic resonance imaging (fMRI), ignoring the structure-function coupling in the brain. In this paper, we introduce a graph signal processing (GSP) based hub detection framework that utilizes both the structural connectivity and the functional activation to identify hub nodes. The proposed framework models functional activity as graph signals on the structural connectivity. Hub nodes are then detected based on the premise that hub nodes are sparse, have higher level of activity compared to their neighbors, and the non-hub nodes' activity can be modeled as the output of a graph-based filter. Based on these assumptions, an optimization framework, GraFHub, is formulated to learn the coefficients of the optimal polynomial graph filter and detect the hub nodes. The proposed framework is evaluated on both simulated data and resting state fMRI (rs-fMRI) data from Human Connectome Project (HCP). △ Less

Submitted 22 October, 2024; originally announced October 2024.

Comments: 13 pages, 4 figures

Ichigo: Mixed-Modal Early-Fusion Realtime Voice Assistant

Authors: Alan Dao, Dinh Bach Vu, Huy Hoang Ha

Abstract: Large Language Models (LLMs) have revolutionized natural language processing, but their application to speech-based tasks remains challenging due to the complexities of integrating audio and text modalities. This paper introduces Ichigo, a mixed-modal model that seamlessly processes interleaved sequences of speech and text. Utilizing a tokenized early-fusion approach, Ichigo quantizes speech into… ▽ More Large Language Models (LLMs) have revolutionized natural language processing, but their application to speech-based tasks remains challenging due to the complexities of integrating audio and text modalities. This paper introduces Ichigo, a mixed-modal model that seamlessly processes interleaved sequences of speech and text. Utilizing a tokenized early-fusion approach, Ichigo quantizes speech into discrete tokens and employs a uniform transformer-based architecture for both speech and text modalities. This method enables joint reasoning and generation across modalities without the need for separate adapters. We present a comprehensive training methodology, including pre-training on multilingual speech recognition datasets and fine-tuning on a curated instruction dataset. Ichigo demonstrates state-of-the-art performance on speech question-answering benchmarks, outperforming existing open-source speech language models and achieving comparable results to cascaded systems. Notably, Ichigo exhibits a latency of just 111 ms to first token generation, significantly lower than current models. Our approach not only advances the field of multimodal AI but also provides a framework for smaller research teams to contribute effectively to open-source speech-language models. △ Less

Submitted 4 April, 2025; v1 submitted 20 October, 2024; originally announced October 2024.