jina-reranker-v3 is a 0.6B parameter multilingual document reranker that introduces a novel last but not late interaction. Unlike late interaction models such as ColBERT that perform separate encoding followed by multi-vector matching, our approach conducts causal self-attention between query and documents within the same context window, enabling rich cross-document interactions before extracting contextual embeddings from the last token of each document. This compact architecture achieves state-of-the-art BEIR performance with 61.94 nDCG@10 while being significant smaller than generative listwise rerankers.

翻译：jina-reranker-v3 是一个拥有 6 亿参数的多语言文档重排序模型，它引入了一种新颖的“最终而非延迟”交互机制。与 ColBERT 等延迟交互模型（先进行独立编码，再进行多向量匹配）不同，我们的方法在同一上下文窗口内对查询和文档进行因果自注意力计算，从而在从每个文档的最后一个词元提取上下文嵌入之前，实现丰富的跨文档交互。这种紧凑的架构在 BEIR 基准测试中取得了 61.94 nDCG@10 的最新最优性能，同时其模型规模显著小于生成式列表重排序模型。