研究人员推出 KaLM-Reranker-V1，这是一种新颖的重排模型，专为大规模检索系统的效率而设计。该模型使用具有 Matryoshka 嵌入池化和交叉注意力的编码器-解码器架构来解耦查询和段落的计算。KaLM-Reranker-V1 有三种尺寸：Nano（0.27B 参数）、Small（1B 参数）和 Large（4B 参数）。在 BEIR、MIRACL 和 LMEB 等基准测试上的实验表明，KaLM-Reranker-V1 取得了有竞争力的性能，其中 Nano 模型甚至可以媲美更大的嵌入模型。 AI
影响 该模型提供了一种更高效的文档重排方法，有望提高信息检索系统的性能和可扩展性。
排序理由 该集群描述了一篇介绍新型 AI 模型的最新研究论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
As retrieval systems scale, high-quality reranking becomes increasingly important. However, most existing rerankers, whether encoder-based or decoder-based, jointly encode the query and passage, tightly coupling their computation and limiting deployment efficiency as well as flex…
As retrieval systems scale, high-quality reranking becomes increasingly important. However, most existing rerankers, whether encoder-based or decoder-based, jointly encode the query and passage, tightly coupling their computation and limiting deployment efficiency as well as flex…
KaLM-Reranker-V1 is a fast reranker that decouples query and passage computation using encoder-decoder architecture with Matryoshka embedding pooling and cross-attention for efficient relevance modeling.