netease-youdao / BCEmbedding

Netease Youdao's open-source embedding and reranker models for RAG products.
Apache License 2.0
1.3k stars 85 forks source link

bce-reranker模型的分数什么普遍偏低 #25

Open shell-nlp opened 5 months ago

shell-nlp commented 5 months ago

我使用 sentence_transformers 进行部署, 输入: [[你是谁,你是谁],[你是谁,今年几岁]] 输出 [0.625,0.425] 为什么第一个 pair 算出的分数这么低

shenlei1020 commented 5 months ago

你好,这是我们算法的核心创新点 - “绝对”分数,既能rerank,又能过滤低质量passage https://github.com/netease-youdao/BCEmbedding/issues/15#issuecomment-1950247627