让视觉文档检索根据查询动态调整页面表示，提升精准度。
Argus-Retriever: Vision-LLM Late-Interaction Retrieval with Region-Aware Query-Conditioned MoE for Visual Document Retrieval
- Query-conditioned Mixture-of-Experts 模块按区域选择专家，使文档表征依赖查询。
- 9B模型在ViDoRe V1上达92.67 NDCG@5，综合榜单86.0，开源模型最高。
- 小规模训练（9%数据）+ 1024维头，兼顾效率与性能，适合集成到智能代理中。
晚交互式视觉语言检索器将每页文档表示为多个视觉标记嵌入，并通过MaxSim评分查询。现有系统如ColPali、ColQwen、ColNomic和Nemotron ColEmbed在生成文档嵌入时未见查询，导致同一页面对表格查询、图表问题和版面敏感证据请求均作相同表示。本文提出 extbf{Argus}，基于Qwen3.5-VL的查询条件晚交互检索器家族。Argus引入区域感知的Mixture-of-Experts模块：查询编码器生成检索嵌入与紧凑上下文向量，文档按空间区域池化，查询感知路由器为每个区域选择潜在专家，再进行MaxSim计算。输出仍为多向量索引，兼容ColPali风格检索，但文档表示变为查询相关（即$/mathbf{D}(q)$）。所有Argus模型使用1024维检索头，低于近期SOTA系统的2560/4096维，且仅用约9%公开监督数据训练。9B模型在ViDoRe V1上达92.67 NDCG@5，综合V1+V2榜单86.0，为当前公开晚交互模型最高分。在ViDoRe V3上，其与Qwen3.6-27B代理管道结合，使NDCG@10从60.28提升至64.80，证明该检索器既可独立使用，也可作为迭代式LLM代理的搜索基础。
原文摘要 · Abstract (English)
Late-interaction vision-language retrievers represent each document page as many visual token embeddings and score queries with MaxSim. In systems such as ColPali, ColQwen, ColNomic, and Nemotron ColEmbed, the document embeddings are produced without seeing the query, so the same page is represented identically for a table lookup, a chart question, and a layout-sensitive evidence request. We introduce \textbf{Argus}, a family of query-conditioned late-interaction retrievers built on Qwen3.5-VL. Argus adds a region-aware Mixture-of-Experts module: the query encoder produces both retrieval embeddings and a compact context vector, the document page is pooled into spatial regions, and a query-aware router selects latent experts per region before MaxSim. The output remains a multi-vector index compatible with ColPali-style retrieval, but the document representation is now dependent on the query (i.e., $\mathbf{D}(q)$). All Argus models use a 1024-dimensional retrieval head, compared with the 2560-dimensional and 4096-dimensional heads of recent state-of-the-art systems, and are trained on roughly 9\% of the available public supervision rather than the full pool. The 9B model reaches \textbf{92.67} NDCG@5 on ViDoRe V1 and \textbf{86.0} NDCG@5 on the combined V1+V2 leaderboard, the highest reported value for an open late-interaction model on the combined leaderboard. Wrapped in a Qwen3.6-27B agentic retrieval pipeline on ViDoRe V3, Argus-9B further improves its NDCG@10 from 60.28 to \textbf{64.80} over public tasks, showing that the same retriever serves both as a strong standalone system and as a search primitive for iterative LLM agents.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。