arXiv cs.CL· Xinping Zhao, Jiaxin Xu, Ziqi Dai, Xin Zhang, Huiyao Chen, Shouzheng Huang, Xianhao Xiong, Danyu Tang, Xinshuo Hu, Guohong Fu, Meishan Zhang, Baotian Hu·· 6 天前AI 评分24
Lychee-Reranker：面向压缩文档重排序的快速非后期交互模型
Lychee-Reranker: Fast but Not Late Interaction for Compressed Document Reranking
AI 导读
研究者提出 Lychee-Reranker，一种解耦查询与文档计算的 FBNL 重排序模型，通过离线编码文档与 cross-attention 建模相关性，提供 Nano、Small、Large 三个版本，激活参数分别为 0.27B、1B、4B。
正文 · 原文
Authors:Xinping Zhao, Jiaxin Xu, Ziqi Dai, Xin Zhang, Huiyao Chen, Shouzheng Huang, Xianhao Xiong, Danyu Tang, Xinshuo Hu, Guohong Fu, Meishan Zhang, Baotian Hu
Abstract:As retrieval systems scale, effective and efficient reranking becomes increasingly important. However, most existing encoder- and decoder-based rerankers jointly process every query--passage pair, tightly coupling their online computation and limiting deployment efficiency and flexibility. We present Lychee-Reranker, a fast but not late-interaction (FBNL) reranker that decouples query and passage computation while retaining expressive relevance modeling. Built on an encoder--decoder architecture, Lychee-Reranker pre-encodes passages using Matryoshka embedding pooling, while its decoder models system and user instructions together with query intent; cross-attention then captures fine-grained relevance between the resulting query context and passage representations. Together, these designs offer four key advantages: (i) efficiency from offline passage encoding, (ii) expressiveness from cross-attention, (iii) compactness from Matryoshka embedding pooling, and (iv) test-time compute through an adjustable compute budget. We instantiate Lychee-Reranker in three sizes, Nano, Small, and Large, with 0.27B, 1B, and 4B activated parameters, respectively. Extensive experiments on BEIR, MIRACL, and LMEB demonstrate strong reranking performance with superior efficiency. On BEIR and MIRACL, our models achieve competitive performance in multi-domain and multilingual reranking, on par with strong industrial rerankers such as the Qwen3/BGE-Reranker series. On LMEB-Dialogue, a compact embedding model paired with our Nano reranker, which has only 0.27B activated parameters, remains competitive with 7--12B embedding models. Lychee-Reranker was previously known as KaLM-Reranker-V1. Data and models are available at this https URL.
| Comments: | Technical Report, 31 pages; |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2606.22807 [cs.CL] |
| (or arXiv:2606.22807v4 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2606.22807 arXiv-issued DOI via DataCite |
Submission history
From: Xinping Zhao [view email]
[v1]
Mon, 22 Jun 2026 03:36:31 UTC (2,597 KB)
[v2]
Tue, 7 Jul 2026 08:14:04 UTC (2,601 KB)
[v3]
Tue, 22 Sep 2026 06:59:06 UTC (5,765 KB)
[v4]
Fri, 9 Oct 2026 06:46:50 UTC (5,405 KB)
来源：arXiv cs.CL · arxiv.org