happy8825/compare_var1_var7

该数据集名为compare_var1_var7，主要用于比较两种变体（var1和var7）在多模态长文档理解任务中的性能。数据集包含训练集，共1073个样本。特征包括多种分数（如var1_score和var7_score）、文档ID和类型、证据和相关页面、问题和答案，以及两种变体的输入和输出。性能评估基于不同的证据来源（如纯文本、图表、表格等）和页面长度（无页面、单页、多页），并提供了每种情况下的准确率。

The dataset named compare_var1_var7 is designed to compare the performance of two variants (var1 and var7) in a multi-modal long document understanding task. It includes a training set with 1073 examples. Features encompass various scores (e.g., var1_score and var7_score), document IDs and types, evidence and relevant pages, questions and answers, and inputs/outputs for the two variants. Performance evaluations are based on different evidence sources (e.g., pure-text, figures, tables) and page lengths (no pages, single page, multiple pages), with accuracy rates provided for each scenario.