Research Interests

The final goal of my research is to advance AI (Artificial Intelligence) systems, as characterized by HAL 9000 in 2001: A Space Odyssey and J.A.R.V.I.S. in Iron Man. They are typically autonomous, interactive agents capable of understanding human intent and making thoughtful decisions. To ensure the trustworthy application in high-stakes domains such as healthcare or autonomous driving, intelligent agents must retrieve relevant evidence from various types of information sources to support reasoning, planning or action. These include extracting domain-specific knowledge from vast and heterogeneous external database (Wikipedia, PubMed, Knowledge Graphes, etc.), accessing most up-to-date information from Web and social media, or recalling useful past experiences stored in their internal memory modules, and so on. Such increasing demand for fast and accurate information access underscores the pivotal role of Information Retrieval (IR) systems in the path towards AGI (Artificial General Intelligence).

Therefore, the core of my research focuses on improving retrieval systems, including both text retrieval and multi-modal retrieval (e.g., image, video, and audio retrieval). Text retrieval serves as a natural starting point, as current large language models (LLMs), despite their transformative impact on many natural language understanding and generation tasks, have not yet led to a fundamental paradigm shift or consistently significant performance gains in retrieval.

Potential entry points for advancing retrieval systems include continual learning, curriculum learning, and the diversification of pretraining data to better cover a wide range of retrieval scenarios and information needs. Building on advances in text retrieval, a natural next step is the development of multi-modal retrieval systems, either through joint training across modalities or through incremental integration of additional modalities.

Consequently, insights and methodologies from natural language processing, multimodal learning, deep learning, and cognitive science are all highly relevant and valuable to my research agenda.

I would be happy to discuss with you any exciting research ideas or potential collaborations. Please feel free to reach out to me via email at yuchen[dot]hui[at]umontreal[dot]ca.