Stefan Ruseti

Automated Extraction of Answer Candidates for Question Generation

GAW-PO: Preference Optimization with Gradient-Aligned Token Weights

Most preference optimization methods, such as Direct Preference Optimization (DPO), apply preference supervision at the response level, although autoregressive language models are optimized token by token. As a result, all tokens in a rejected response contribute to the negative training signal, including tokens that m...

RoCulturaMCQ: Building a Benchmark While Learning Statistics

A pilot project in which students in a statistics course within a data science engineering program created culturally diverse multiple-choice questions, generated answers using LLMs, and applied statistical methods to assess model accuracy is presented, supporting a cultural injection hypothesis.