Blaise Cruz

Mabuhay! 👋

I’m a PhD student at MBZUAI supervised by Dr. Alham Fikri Aji working on novel approaches to Modeling Multilinguality. In particular, my work centers on methods that reframe multilinguality in an efficient and linguistically-motivated manner, beyond simply cramming hundreds of unique languages within the same billion-ish parameters.

Besides this, I’ve worked on various other topics under the multilinguality and low-resource umbrella, including:

• Code Switching – Multilingual speakers naturally code-switch in two or more languages when speaking to peers, but multilingual models are still lacking in capabilities to understand and execute this phenomenon.
• Resources & Evaluation – More data is often the best remedy to “very little data”. Beyond Filipino resources and benchmarks for my home region, I have also collaborated for Southeast Asian tools and resources as a co-founder of SEACrowd and ACL SIGSEA.
• Beyond Southeast Asia – Beyond my home region, I’ve also worked on global efforts for cultural QA, cultural MT and large-scale cultural benchmarking. More recenty, I’ve dabbled in topics such as theory of mind and competitive programming, leveraging my past as an ICPC contestant.
• Applications in Low-resource – Lastly, I’ve worked on improving performance in tasks such as Multilingual Translation, Question Generation, Fake News Detection, and more – all constrained under low-resource settings.

Prior to my PhD, I was Lead Research Engineer at Samsung Research where I worked on low-resource machine translation and dialogue generation. I’ve also previously been affiliated with Mila - Quebec AI Institute and McGill University, the University of the Philippines, De La Salle University, and Senti AI.

If you’re interested in collaborating or if you want to chat about low-resource languages, feel free to get in touch! You may reach me through my email me (at) blaisecruz (dot) com.