← 研踪 PaperTrailarXiv:2609.02486 · v3
ViSAR: Training-Free Adaptive-$k$ Retrieval for Document Visual Question Answering
首次提交 2026-09-02 19:54
版本更新 2026-11-16 21:12
雷达观测 2026-11-17 17:11
摘要
Document Visual Question Answering (DocVQA) often leverages Retrieval-Augmented Generation (RAG), where late-interaction encoders are commonly used to identify document pages relevant to a user query, before answer generation by a Large Vision-Language Model (LVLM). Existing approaches typically retrieve a fixed top- number of pages regardless of query complexity, which increases LVLM latency and may degrade answer accuracy when fewer pages would be enough. We introduce ViSAR (Visual Semantic Activation Retrieval), a training-free, budget-aware adaptive- retrieval method for late-interaction visual document retrieval. ViSAR operates directly in the embedding space to construct a query-conditioned page-level similarity matrix that highlights query-relevant semantics and dynamically determines the number of pages to retrieve, within a given page budget. We evaluate our method across multiple datasets, encoders, and LVLMs. ViSAR retrieves compact, query-adapted page sets that reduce end-to-end RAG latency by up to 58.9\%, with no statistically significant accuracy loss and significant gains in several settings, compared with fixed top- and adaptive retrieval heuristics. Furthermore, we show that the similarity matrix structure correlates with answer accuracy, suggesting future directions for retrieval quality-aware document understanding.
作者备注:32 pages, 20 figures, 10 tables
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v3 · 2026-11-16 21:12 · 最新观测版本
ViSAR: Training-Free Adaptive-$k$ Retrieval for Document Visual Question Answering
Document Visual Question Answering (DocVQA) often leverages Retrieval-Augmented Generation (RAG), where late-interaction encoders are commonly used to identify document pages relevant to a user query, before answer generation by a Large Vision-Language Model (LVLM). Existing approaches typically retrieve a fixed top- number of pages regardless of query complexity, which increases LVLM latency and may degrade answer accuracy when fewer pages would be enough. We introduce ViSAR (Visual Semantic Activation Retrieval), a training-free, budget-aware adaptive- retrieval method for late-interaction visual document retrieval. ViSAR operates directly in the embedding space to construct a query-conditioned page-level similarity matrix that highlights query-relevant semantics and dynamically determines the number of pages to retrieve, within a given page budget. We evaluate our method across multiple datasets, encoders, and LVLMs. ViSAR retrieves compact, query-adapted page sets that reduce end-to-end RAG latency by up to 58.9\%, with no statistically significant accuracy loss and significant gains in several settings, compared with fixed top- and adaptive retrieval heuristics. Furthermore, we show that the similarity matrix structure correlates with answer accuracy, suggesting future directions for retrieval quality-aware document understanding.
查看此版本解读 →查看此版本原文 ↗