MS-Bench
收藏DataCite Commons2025-05-14 更新2025-05-17 收录
下载链接：
https://dataverse.harvard.edu/citation?persistentId=doi:10.7910/DVN/MKRTMN
下载链接
链接失效反馈官方服务：
资源简介：
<p>
This is <strong>MS-Bench</strong>, the first comprehensive benchmark co-developed with archaeologists,<br>
comprising <strong>5,076 high-resolution images</strong> from <em>4th to 14th century</em> and
<strong>9,982 expert-curated questions</strong> across nine sub-tasks aligned with archaeological workflows.
</p>
<p>
Through four prompting strategies, we systematically evaluate <strong>32 LMMs</strong> on their:
</p>
<ul>
<li>effectiveness</li>
<li>robustness</li>
<li>cultural contextualization</li>
</ul>
提供机构：
Harvard Dataverse
创建时间：
2025-05-14
