Дослідження здатності малої мовної моделі до міркування

Abstract

The capability of the small language model to solve reasoning tasks was tested and evaluated. Experiments were conducted with zero-shot and few-shot prompting, after applying Supervider Fine-Tuning to TinyLlama small language model. The results of the experiments were presented and analyzed, along with proposed potential ways to improve the model's accuracy in solving reasoning tasks.