InSight-doc：面向長文檔理解的代理式視覺感知

摘要

長文件理解通常需要在許多視覺豐富的頁面上進行推理，導致推論成本高昂且容易發生上下文衰減。在這項工作中，我們提出 InSight-doc，一個代理式視覺感知框架，將視覺解析度視為自適應的推理時間資源。InSight-doc 從低解析度開始，選擇性地放大到高解析度區域以獲取更精細的證據，且不依賴任何外部檢索器。為了訓練這樣的代理，我們建構了一個主動感知語料庫，包含 17.9K 個高品質 SFT 範例，帶有區域級別的放大軌跡，並附有 19.2K 個高難度 RL 範例。透過 SFT+RL，InSight-doc-8B 在文件 VQA 基準上將基線提升了 4.3 至 16.4 個準確度百分點。在長文件上，它將幻覺減少了超過 40%，並將推論延遲降低了 41% 至 68%，同時保持準確度領先。我們的程式碼、資料集和模型已公開於 https://github.com/m-Just/InSight-doc 。

English

Long-document understanding often requires reasoning over many visually rich pages, making inference costly and prone to context rot. In this work, we propose InSight-doc, an agentic visual perception framework that treats visual resolution as an adaptive reasoning-time resource. InSight-doc starts from low resolution and selectively zooms into high-resolution regions for finer evidence, without relying on any external retriever. To train such an agent, we construct an active-perception corpus of 17.9K high-quality SFT examples with region-level zoom-in trajectories, accompanied by 19.2K hard RL examples. Through SFT+RL, InSight-doc-8B improves the baseline by 4.3--16.4 accuracy points over document VQA benchmarks. On long documents, it reduces hallucination by more than 40% and inference latency by 41%--68% while maintaining an accuracy lead. Our code, datasets, and model are released at https://github.com/m-Just/InSight-doc .