Vector Discovery Benchmark 75 Result
收藏Zenodo2025-09-05 更新2026-05-26 收录
下载链接：
https://zenodo.org/doi/10.5281/zenodo.17064743
下载链接
链接失效反馈官方服务：
资源简介：
This is a benchmark result is made up of 75 natural language queries to cover a range of retrieval challenges. We developed the benchmark result to assess the viability for discovery of multi-modal vector embedding. See accompanying paper.
The queries are organized by category from simple text lookups to complex multimodal reasoning tasks. The dataset contains each query, the ground truth defined for that query, and the top-5 results returned for each of the distance measures used (cosine similarity, dot product, euclidean distance, and manhattan distance).
The actual digitized pages are not included in this result set as we do not have permission to share them.
提供机构：
Zenodo创建时间：
2025-09-05
