🎓 I am a PhD student @ MBZUAI, focusing on multimodal alignment, LLM, VLM, and evaluation.
💡 My research interests include:
- Multimodal Imbalance: I believe that imbalanced learning is a significant bottleneck that prevents us from obtaining reliable multimodal models, as modality shortcuts and biases can harm both performance and the objectivity of evaluation. My work focuses on discovering its root causes and exploring methods to better align models to prevent such issues.
- LLM/VLM Alignment: I also work on both architectural and non-architectural adaptations (knowledge enrichment, data reformulation, RL) to address above issues and/or improve multimodal language modeling in general.
- Large-Scale Evaluations: I often question model robustness in scenarios with varying resource levels; however, probing this requires designing both broad and specific evaluation coverage. My work in this area aims to design benchmarks that assess the inclusivity of multimodal models, specifically by addressing concept underrepresentation through targeted data curation in multilingual and multicultural domains.
🇸🇬 I was also a Research Engineer at SMU 🇸🇬, advised by Prof. Chong-Wah Ngo, working on the intersection of multimodal and multilingual learning.
🧪 Previously, I earned my CS degree at ITB 🇮🇩 under Prof. Ayu Purwarianti, working on explainable synthetic data generation.
📬 Email me • 🌐 Website • 📝 Google Scholar • 💼 LinkedIn