Andre Niyongabo Rubungo
I am a 5th-year PhD candidate in Computer Science at Princeton University, advised by Prof. Adji Bousso Dieng. My research interests are large language models and AI for Science, with a focus on leveraging large language models for scientific discovery.
My research internships at IBM Research span LLM post-training (2025) and agentic coding models (2026). Previously, I worked on low-resource machine translation at Polytechnic University of Catalonia, Barcelona Tech (UPC) under the supervision of Prof. Marta R. Costa-jussà and Dr. Carlos Escolano and interned with Meta AI’s NLLB team under the supervision of Angela Fan. I earned my bachelor’s and master’s degrees in Computer Science at the University of Electronic Science and Technology of China (UESTC).
Research
My work connects three themes:
- Language models for materials discovery. Developing representations and models to predict crystal properties and metal–organic framework free energies (LLM-Prop, MOF free energy prediction).
- Evaluating scientific prediction. Building benchmarks and testing how well models generalize to materials beyond their training data (LLM4Mat-Bench, out-of-distribution generalization).
- Language modeling and retrieval for low-resource languages. Using linguistic structure to improve language representations and retrieval-augmented generation (KinyaBERT, KinyaColBERT).
Selected Publications and Preprints
See all publications for the full list.
-
Highly Accurate and Fast Prediction of MOF Free Energy Via Machine Learning
Journal of the American Chemical Society (JACS), 2025 [Paper] [Code]
-
LLM-Prop: predicting the properties of crystalline materials using large language models
-
LLM4Mat-Bench: Benchmarking Large Language Models for Materials Property Prediction
Machine Learning: Science and Technology, 2024 [Paper] [Code]
-
Probing Out-of-distribution Generalization in Machine Learning for Materials
-
KinyaColBERT: A Lexically Grounded Retrieval Model for Low-Resource Retrieval-Augmented Generation
-
KinyaBERT: A Morphology-aware Kinyarwanda Language Model
Languages: Kinyarwanda (native); English and Mandarin Chinese (fluent); French and Kirundi (intermediate); Spanish (beginner).