Nguyen Quang Huy

Research Intern @ NLP & KD Lab, TDTU, Vietnam

B.Sc. Computer Science student at Ton Duc Thang University and Research Intern at the NLP & Knowledge Discovery Lab. My interests span Computer Vision, Multimodal AI, Large Language Models, and AI systems.

Mobile version coming soon

Nguyen Quang Huy

Research Intern @ NLP & KD Lab, TDTU, Vietnam

A lightweight mobile experience is on the way. For now, you can still reach me here.

About Me

I am a third-year Computer Science student at Ton Duc Thang University and a Research Intern at the NLP & Knowledge Discovery Lab. My interests include Computer Vision, Multimodal AI, Image Forensics, Large Language Models (LLMs), and AI systems.

My work spans both research and engineering. I have developed AI-generated image detection methods, built retrieval-augmented generation (RAG) pipelines, LLM-powered chatbots, and AI agent systems, while also contributing to benchmark construction and domain-specific AI applications.

I am the author of peer-reviewed research publications in AI and image forensics, and have received awards in scientific research and academic competitions. I am currently seeking AI Research and AI Engineer internship opportunities.

I am always open to research collaborations, AI projects, and interdisciplinary initiatives. If you are interested in working together, please feel free to reach out through the Contact section below.

I am also actively pursuing future Master's degree opportunities in the United States, United Kingdom, South Korea, Japan, China, and Australia. I am particularly interested in fully funded graduate programs that provide 100% tuition coverage and living expenses support. If your research group or laboratory is recruiting prospective graduate students, research assistants, or Master's candidates, I would be delighted to discuss potential opportunities.

Research output

FALCON: Forensic-Aware Language-Guided Contrastive Learning for Generalized Synthetic Image Detection

Under review - [FITAT 2026] 18th International Conference on Frontiers of Information Technology, Applications and Tools

Nguyen, Q. H., & Pham, V. H. (2026)

Recent advances in generative models have significantly improved the realism of synthetic images, making cross-generator synthetic image detection increasingly challenging. Existing forensic detectors often learn generator-specific artifacts and therefore suffer substantial performance degradation when evaluated on unseen architectures. In this paper, we propose FALCON (Forensic-Aware Language-guided Contrastive Learning), a language-guided forensic framework that uses forensic-aware textual supervision to learn transferable visual representations. FALCON aligns image features with textual descriptions of semantic content and forensic traces, and then refines the detector with a hybrid contrastive and classification objective. To evaluate cross-generator robustness, we construct UniRF-112K, a balanced benchmark of 112,000 real and synthetic images spanning GANs, diffusion models, transformer-based generation, and flow matching. Under a one-generator training protocol, models are trained on ProGAN and evaluated across diverse held-out generators. Experimental results show that the best FALCON variant achieves a mean Average Precision (mAP) of 83.30%, outperforming LASTED by 8% and obtaining strong gains on several challenging unseen generators, including StyleGAN3, DiT, and Flux.1. Ablation results further indicate that combining image-class context with forensic trace descriptions provides the most effective textual supervision for generalized synthetic image detection.

ASBW: A Frequency-Domain Analysis Approach for Distinguishing GAN-Generated Images from Real Images

[DCEST 2026] International Conference on Digital Convergence in Economics, Society and Technology, pp. 95-105

Nguyen, Q. H., & Pham, V. H. (2026)

The rapid advancement of Generative Adversarial Networks (GANs) continues to challenge digital media integrity, while purely spatial-domain forensic models remain sensitive to overfitting and weakly generalizable artifacts. To address this, we propose the Adaptive Spectral-Band Weighting (ASBW) layer, a lightweight frequency-domain module that applies the Discrete Fourier Transform (DFT) and a compact Multi-Layer Perceptron (MLP) to reweight discriminative spectral cues caused by GAN generation. We evaluate ASBW insertion at different positions across the ResNet-18 pipeline (after each convolutional block), and compare these placement strategies with the baseline CNN. Results on the real/fake face benchmark show that ASBW effectiveness is strongly position-dependent: the full configuration achieves the best performance (93.12%), while placing ASBW after later blocks (e.g., Position 3: 92.22%) consistently improves over the baseline (89.77%). These findings highlight that frequency-aware amplification is most beneficial when integrated with suitable feature hierarchy depth, and provide practical guidance for architecture design in GAN image forensics.

Academic & research timeline

Jun 2026 - Present

Research Intern

  • Fine-tune pretrained LLMs and adapt multimodal models for domain-specific tasks.
  • Curate, preprocess, and annotate datasets to support experiments and benchmarks.
  • Develop retrieval-augmented systems and prototype LLM-based applications.
Sep 2023 - Present

B.Sc. Computer Science

  • AI & Machine Learning: Introduction to AI, Machine Learning.
  • Deep Learning & Multimodal AI: Deep Learning, Natural Language Processing, Computer Vision.
  • Data Science & Knowledge Discovery: Knowledge Discovery and Data Mining, Mining Massive Datasets.
  • Research & Industry Training: Industrial Experience Project, Graduation Internship, Graduation Thesis.
Sep 2020 - May 2023

High School Diploma

  • 3rd Prize, Provincial Excellent Student Competition, Lam Dong
  • Member, Lam Dong Provincial Team for National Excellent Student Competition

Let's connect

I'm currently open to AI Research and AI Engineer internship opportunities in Ho Chi Minh City. If you're working on Multimodal AI, Image Forensics, LLMs, RAG, or AI Agent systems, I'd love to connect. I also welcome research collaborations and am actively seeking fully funded Master's opportunities with research groups and laboratories worldwide.