Nhi N. Truong

Stanford University

Stanford, California

Hi, I’m Nhi. I love building and evaluating machine learning systems. I currently work with Professor Sanmi Koyejo at the Stanford Trustworthy AI Research Lab, where I help develop evaluation frameworks for large language models, and this summer I’m joining Stanford HAI as a research intern. My work has been published at NAACL and ICLR, and featured in The New York Times and the HAI news room.

I study Computer Science and Economics at DePauw University, where I’m an Honor Scholar, a Management Fellow, a Science Research Fellow, and part of the Information Technology Associates Program.

Previously, I worked on time series forecasting at the VNU-HCMUT AI Lab and contributed to The R Project through Google Summer of Code.

My interests are in machine learning, natural language processing, and time series forecasting, especially their applications in healthcare and education. Feel free to reach out.

Selected Projects

DePause

A campus mental health app that turns 5-second mood check-ins into a live, anonymous map of how the campus feels — 2nd place at TigerHacks 2026.

Selected Publications

NAACL

Crossing Linguistic Horizons: Finetuning and Comprehensive Evaluation of Vietnamese Large Language Models

Sang T. Truong^*, Duc Q. Nguyen^*, Toan Nguyen^*, and 4 more authors * Equal contribution.

In Findings of the Association for Computational Linguistics: NAACL 2024. Featured in The New York Times and Stanford HAI. , Mar 2024

Abs HTML PDF

Recent advancements in large language models (LLMs) have underscored their importance in the evolution of artificial intelligence. However, despite extensive pretrained on multilingual datasets, available open-sourced LLMs exhibit limited effectiveness in processing Vietnamese. The challenge is exacerbated by the absence of systematic benchmark datasets and metrics tailored for Vietnamese LLM evaluation. To mitigate these issues, we have finetuned LLMs specifically for Vietnamese and developed a comprehensive evaluation framework encompassing 10 common tasks and 31 metrics. Our evaluation results reveal that the fine-tuned LLMs exhibit enhanced comprehension and generative capabilities in Vietnamese. Moreover, our analysis indicates that models with more parameters can introduce more biases and uncalibrated outputs and the key factor influencing LLM performance is the quality of the training or fine-tuning datasets. These insights underscore the significance of meticulous fine-tuning with high-quality datasets in enhancing LLM performance.
ICLR WS

Hybrid Transformer and Holt-Winter’s Method for Time Series Forecasting

Nhi N. Truong, Duc Q. Nguyen, Jeffrey Gropp, and 1 more author

In Time Series for Health Workshop, International Conference on Learning Representations (ICLR), Mar 2024

Abs HTML PDF

Time series forecasting is an important research topic in machine learning due to its prevalence in social and scientific applications. Multi-model forecasting paradigm, including model hybridization and model combination, is shown to be more effective than single-model forecasting in the M4 competition. In this study, we hybridize exponential smoothing with transformer architecture to capture both levels and seasonal patterns while exploiting the complex non-linear trend in time series data. We show that our model can capture complex trends and seasonal patterns with moderately improvement in comparison to the state-of-the-arts result from the M4 competition.