For records in your dataset where {{ issue.domain }},
we found a significantly higher number of overconfident wrong predictions
({{ issue.info.fail_idx|length }} samples, corresponding to
{{ (issue.info.metric_value_slice * 100)|round(1) }}% of the wrong predictions in the data slice).
{{ issue.examples(3).to_html(justify="left", notebook=True) | replace("\\n", "
") | safe }}