Jina 发布轻量重排序模型 v3.5，混合注意力与自蒸馏实现性能突破
原标题：jina-reranker-v3.5: Faster Listwise Reranking with Hybrid Attention and Self-Distillation
AI 摘要
Jina AI 发布 jina-reranker-v3.5 重排序模型，参数仅 0.6B，性能超越 1.5B 的 mxbai-rerank-large-v2 和 4B 的 Qwen3-Reranker-4B，在 BEIR 上达到 63.20，半结构化检索提升 9.6 nDCG@10。该模型采用混合注意力（3L2G）和自蒸馏技术，在保持质量的同时显著提升推理速度。
正文节选
Qwen3-Reranker-4B with roughly 7× fewer parameters, and it reranks up to 1.56× faster than v3 on long documents. Its biggest jump is on semi-structured retrieval: +9.6 nDCG@10 over v3 on field-constrained records. Three changes get us there. A hybrid attention schedule that replaces most global layers with sliding windows while pinning the terminal layer to global. A training mixture curated from the failure modes of legal, medical, financial, multilingual, and structured retrieval. And a three-