We employ the original
Llama-3-8B-Instruct model, which we further fine-tune using carefully selected samples from our compiled dataset (available at
https://github.com/rhyang2021/LoGU). This refined model is subsequently utilized within our LoGU task, specifically targeting the challenges of
uncertainty-suppression and
uncertainty-misalignment. For additional details, please refer to our
Github and
Paper.
