Chart-HQA
收藏资源简介:
Chart-HQA是一个由浙江大学和阿里巴巴集团合作构建的开放域图表假设性问题回答数据集。该数据集通过人类与大型语言模型互动的方法,生成具有多样性和高质量的假设性问题。它包含了2173个假设性问题,覆盖了900种指令提案和947个图表,旨在测试模型在图表内容理解方面的零样本推理能力。数据集中的假设性问题格式为开放词汇,要求对底层图表数据进行反事实操作,这为图表理解任务带来了新的挑战。
Chart-HQA is an open-domain chart hypothetical question answering dataset jointly constructed by Zhejiang University and Alibaba Group. This dataset generates diverse and high-quality hypothetical questions via the interaction between humans and large language models. It contains 2173 hypothetical questions, covering 900 instruction proposals and 947 charts, aiming to test the zero-shot reasoning ability of models when understanding chart content. The hypothetical questions in the dataset adopt an open-vocabulary format, requiring counterfactual operations on the underlying chart data, which introduces new challenges to the chart understanding task.




