How should we prepare the element scores in the submitted json files? Are they original logits from the models, or the binarized results (0 or 1)?
Posted by: duanxiongwa @ Feb. 10, 2025, 2:35 a.m.We used the accuracy to evaluate the results and therefore used the binary results, where >= 0.5 annotated data is considered to be 1
Posted by: hsh @ Feb. 19, 2025, 1:21 p.m.