Alphazero Based Neural Networks for 2-Move Chess
Advisors/Collaborators: Professor Stoyan Dimitrov
Trained a series of neural networks for modified chess rulesets using a similar approach as Alphazero.
Seeking to solve 2-move chess with optimal computer play.
Large Language Models for Scientific Problem Solving Based on Retrieval Augmented Hint Generation
Advisors/Collaborators: Professor Yongfeng Zhang, Professor Alexandre Morozov, PhD Student Haoming Gong
Analyzed how LLMs handle reasoning tasks when given a "problem -> algorithm -> application" pipeline.
Aggregated a large dataset of problems and algorithms for fine tuning and testing.
Studied how larger language models could improve responses of smaller language models.
Large Language Models for Synthetic Data Generation
Advisors/Collaborators: Professor Tomasz Imielinski
Study on using LLMs for data generation, storage, and compression.
Generated simulation stock market datasets for "realism" accuracy evaluations.
Tracked how LLMs detected patterns within large datasets.
