Research Portfolio
Advancing artificial intelligence research with a focus on explainability, interpretability, and real-world applications for mental health and social impact.
Publications & Presentations
Published research, conference presentations, and ongoing contributions to AI and mental health
Featured Publications
Adversarial Attacks on Reinforcement Learning-based Medical Questionnaire Systems
arXiv:2508.05677 • Mentored by Prof. Hanan Hibshi, Carnegie Mellon University
Novel research on adversarial robustness in medical AI systems, focusing on input-level perturbation strategies and medical constraint validation. Click to view paper →
Competition & Conference Presentations
🏆 EASRL-MH: Enhanced Adaptive Symptom Reasoning Learning for Mental Health
China International College Students' Innovation Competition (CICSIC) 2025 • Selected as Top 20 in Shaanxi Province
Conference presentation slides Click to view →
Research poster presentation Click to view →
🧠 BERT-Based Emotion Diagnosis Support System with LIME Explanations
China Talented Youth Program National Conference • Huazhong University of Science and Technology • December 2024
Full research paper on emotion diagnosis with AI interpretability Click to view →
Conference poster presentation Click to view →
Research Papers & Surveys
Interpretable Artificial Intelligence: A Survey of Methods and Applications
Comprehensive survey of interpretable AI methods and their applications across domains. Click to view paper →
Public Engagement & Science Communication
"Taking a Breath" - Mental Health Awareness Post on Burnout Prevention
Featured on @ScholarsOfPioneer• Research-backed mental health guidance
Science communication piece addressing academic burnout, featuring evidence-based strategies for student mental health. Combines psychological research with accessible advice on active recovery, growth mindset, and stress management. Includes 10 peer-reviewed citations and practical applications of psychological research. Click to view post →
Current Research
AI4Health
EASRL-MH: Enhanced Adaptive Symptom Reasoning Learning for Mental Health.
Xi'an Jiaotong University | 2023 - Present
Research mainly focused on AI for healthcare, especially for mental health.
Developed EASRL-MH: Enhanced Adaptive Symptom Reasoning Learning for Mental Health.
This is a project intended to assist mental health diagnosis through reinforcement learning.
Through selecting the most informative questions to ask, the model can diagnose the patient's mental disorder with high accuracy and relatively low number of questions.
Our model has achieved state-of-the-art results on the NCS-R benchmark dataset,
successfully shortening the length of interview to an average of 9.6 questions with 86.4% accuracy and an AUC-ROC of 0.92.
Key Contributions:
- Reinforcement learning for mental health diagnosis
- State-of-the-art results on the NCS-R benchmark dataset
Explainable AI (XAI)
A BERT-Based Emotion Diagnosis Support System with Improved Explainability
Northwestern Polytechnical University,on behalf of China Talented Youth Project | 2024 - 2025
Developed an AI emotional classification system based on BERT. The key innovation is the application of Local Interpretable Model-Agnostic Explanations (LIME) to improve the explainability of the model. The model achieved over 95% classification accuracy on the Sentiment140 dataset consisting over 160 million tweets.
Research Focus:
- BERT-based emotion classification
- LIME for explainability
Educational AI
AI-Powered Educational Tools
Personal Research Project | 2024 - Present
Leveraging hands-on experience teaching 350+ elementary students to inform AI-powered educational research. This unique intersection of practical pedagogy and machine learning is producing innovations in adaptive assessment, personalized learning pathways, and intelligent code review systems tailored for young learners.
Innovation Areas:
- Adaptive assessment systems that adjust to student learning pace
- AI-powered code review and feedback for beginner programmers
- Natural language interfaces for programming education
- Gamification strategies enhanced by machine learning
Research Philosophy
How I approach problems and create meaningful impact
Impact-Driven Research
Every research project I undertake starts with a real-world problem. From developing adversarial-robust medical questionnaire systems to creating emotion diagnosis tools, my work demonstrates how rigorous research can directly improve lives through better healthcare access, enhanced educational tools, and more trustworthy AI systems.
Explainability First
My published research consistently emphasizes explainability - from LIME explanations in emotion diagnosis to interpretable AI surveys. I believe that for AI to be truly beneficial in healthcare and education, users must understand not just what the system predicts, but why it makes those predictions.
Interdisciplinary Collaboration
My research thrives on interdisciplinary collaboration - from working with Carnegie Mellon faculty on adversarial attacks to combining psychological insights with NLP for emotion diagnosis. This collaborative approach ensures technical innovations address genuine human needs across healthcare, education, and beyond.
Ethical AI Development
My work on adversarial robustness and privacy-preserving techniques in medical AI demonstrates a commitment to ethical development. I believe responsible AI means not just building systems that work, but ensuring they're secure, fair, and beneficial for all users - especially in sensitive applications like mental health and education.
Future Research Directions
Looking ahead to the next frontiers in AI research and education
Adversarial Robustness in Healthcare AI
Building on my published arXiv research to develop more secure and robust AI systems for medical applications, with focus on real-world deployment considerations.
Explainable AI for Education
Extending my interpretable AI research to create educational tools that can explain their reasoning to both teachers and students, fostering trust and understanding.
Cross-Cultural Mental Health AI
Developing culturally-aware emotion diagnosis systems that can adapt to different cultural expressions of mental health, building on my BERT-based emotion research with international collaboration.