RLinf/WideSeek-R1-Corpus

WideSeek-R1语料库是一个用于训练WideSeek-R1模型的数据集，包含三个主要部分：wiki_corpus.jsonl（作为模型的搜索工具，根据查询返回最相关的片段）、wiki_webpages.jsonl（作为模型的访问工具，根据特定URL返回完整的网页内容）和qdrant/（一个本地Qdrant向量数据库，用于高效检索并作为搜索工具的核心后端）。这些数据来源于ASearcher-Local-Knowledge数据集。The WideSeek-R1 Corpus is a dataset used to train the WideSeek-R1 model, consisting of three main components: wiki_corpus.jsonl (serves as the models Search tool, returning the most relevant snippets given a query), wiki_webpages.jsonl (serves as the models Access tool, returning the full webpage content given a specific URL), and qdrant/ (a local Qdrant vector database built for efficient retrieval, acting as the core backend for the Search tool). Both wiki_corpus.jsonl and wiki_webpages.jsonl are sourced from the ASearcher-Local-Knowledge dataset.