quicktensor/blockrank-beir-evals
收藏Hugging Face2025-11-06 更新2025-11-15 收录
下载链接:
https://hf-mirror.com/datasets/quicktensor/blockrank-beir-evals
下载链接
链接失效反馈官方服务:
资源简介:
ICR-BEIR-Evals是一个专门为上下文中的文档排名任务设计的评估数据集,它从BEIR基准派生而来,包含了28,759个查询,跨越11个不同的领域,每个查询都配有了使用Contriever密集检索模型检索到的100个候选文档。这个数据集适用于评估在检索到的候选集合上操作的列表式排名方法。
ICR-BEIR-Evals is an evaluation dataset specifically designed for document ranking tasks in context, derived from the BEIR benchmark. It includes 28,759 queries across 11 diverse domains, each paired with 100 candidate documents retrieved using the Contriever dense retrieval model. This dataset is particularly useful for evaluating listwise ranking approaches that operate on the retrieved candidate sets.
提供机构:
quicktensor



