eval-gemini-2.5-pro-multimed-hard-20260408-1933
收藏资源简介:
该数据集包含 Whisper 模型在 gemini-2.5-pro 上的评估结果。评估使用的数据集为 Trelis/multimed-hard,模型为 google/gemini-2.5-pro。数据集提供了详细的评估指标,包括词错误率(WER)和字符错误率(CER)。数据列包括音频样本(如果源数据集提供)、参考转录文本、模型预测文本、每个样本的 WER 和 CER,以及实体注释和每个样本的实体 CER(如果没有实体则为 -1.0)。此外,数据集还提供了按类别(如解剖学、生物标志物、条件、药物、组织和程序)划分的实体 CER 详细统计。整体实体 CER 为 16.70%。
This dataset contains the evaluation results of the Whisper model on google/gemini-2.5-pro. The evaluation was conducted using the Trelis/multimed-hard dataset, with the model being google/gemini-2.5-pro. The dataset provides detailed evaluation metrics, including Word Error Rate (WER) and Character Error Rate (CER). The data columns include audio samples (if provided by the source dataset), reference transcriptions, model predictions, WER and CER for each sample, entity annotations, and entity CER for each sample (set to -1.0 if no entities are present). In addition, the dataset also provides detailed statistics of entity CER categorized by classes such as anatomy, biomarkers, conditions, drugs, tissues and procedures. The overall entity CER is 16.70%.




