Nazarii Shportun

The authors created ImmigrationQA, a large source-grounded QA dataset for U.S. immigration law, and fine-tuned a small language model (Llama 3.2 3B) on it, achieving a significant performance boost over base models.