Skip to content
View Sv-Arcode's full-sized avatar

Block or report Sv-Arcode

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Sv-Arcode/README.md

Hi, I'm Muyang Chen

Final-year B.Eng. student in Electrical and Computer Engineering (Computer Engineering stream) at The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen), and undergraduate researcher in the Network and Artificial Intelligence System (NAIS) Lab, advised by Prof. Fangxin Wang.

Major GPA 3.90/4.00 (3/60, top 5% in the Computer Engineering stream) · Dean's List 2023–24, 2024–25, 2025–26

🔍 Research Interests

Multimodal large language models · hidden-intent & emotion understanding · multimodal reasoning · model interpretability

📌 Selected Work

  • MISID — Hidden-Intent Recognition in Strategic Deception Games Multimodal, multi-turn benchmark with 3,962 temporally aligned audiovisual segments and a two-tier annotation scheme. Designed FRACTAM's modal decoupling and text reconstruction component, independently handled the inter-annotator agreement study, and contributed to evaluation, ablation studies, and the LLM-as-a-Judge protocol. Co-first author — submitted to AAAI 2027 (under review).
  • AffectiveArt Challenge 2026 — ACM Multimedia Grand Challenge, Track 2 Led a 3-member team; designed the overall architecture; fine-tuned Gemma, Qwen, SigLIP2, and DINOv3 classifiers; and designed and implemented validation-weighted voting that outperformed the best individual model. Independently fine-tuned SigLIP2 and DINOv3 for visual-attribute salience prediction; salience and classification outputs conditioned a LoRA-fine-tuned Qwen3.5-27B description model. Rank 5 / 14 on the official leaderboard (overall score 0.8236).
  • Butterfly-CNN-Interpretability — repo 4-layer CNN classifying 10 visually similar butterfly species (83.23% test accuracy), with Class Activation Mapping implemented from scratch and PCA feature-space visualization. ECE4513 Image Processing and Computer Vision final project.

🏃 Ongoing

  • Preparing PhD applications (Fall 2027) — multimodal reasoning & social intelligence
  • MISID is under review at AAAI 2027

🛠 Skills

Python C++ PyTorch OpenCV Git Linux LaTeX

Popular repositories Loading

  1. Sv-Arcode.github.io Sv-Arcode.github.io Public

    HTML

  2. butterfly-cnn-interpretability butterfly-cnn-interpretability Public

    Butterfly species classification with a CNN — feature maps, PCA and Class Activation Mapping (CAM) implemented from scratch to interpret what the network learns. (IPCV25 final project)

    Jupyter Notebook

  3. Sv-Arcode Sv-Arcode Public

    Undergraduate researcher in AI at CUHK-Shenzhen - multimodal LLMs, intent & emotion understanding, model interpretability

  4. emoart-multitask emoart-multitask Public

    Fine-tune SigLIP2 & DINOv3 for joint arousal / valence / emotion classification on 5,532 annotated artworks (EmoArt-5k), with a shared split, imbalance-aware losses and full eval pipeline.

    Python

  5. fab-g-salience-classifier fab-g-salience-classifier Public

    Multi-label classifier predicting which formal attributes (composition/color/line/brushstroke/light) of an artwork are salient emotion cues - FAB-G Salience Extension

    Python