全局地圖與局部視角下的多視角3D推理密集獎勵

摘要

多視角3D視覺問答（MV3D-VQA）需要將部分觀測整合為連貫的3D場景表示，並選擇具資訊價值的視角進行多步驟空間推理。然而，當前的多模態大型語言模型通常以稀疏的答案層級監督進行訓練，這往往導致跨視角推理不一致且視角選擇脆弱。我們提出DR-MV3D（為MV3D-VQA設計的稠密獎勵），這是一種基於地圖建構的學習框架，提供稠密且可驗證的獎勵來監督推理過程。我們的方法將MV3D-VQA分解為：(i) 異我中心的全域地圖建構，(ii) 問題條件下的視角軌跡規劃，以及(iii) 自我中心的接地推理以進行答案預測。為了讓中間步驟無需人工標註即可學習，我們引入了兩種獎勵：全域一致性獎勵，用於將預測地圖與來自凍結3D視覺基礎模型（例如VGGT + SAM3）的幾何一致偽目標對齊；以及局部軌跡獎勵，用於監督有序視角選擇。我們通過軌跡層級的策略優化（GRPO）來優化完整流程。在MindCube、VSI-Bench和BLINK (MV)上的實驗結果顯示，DR-MV3D在多強基線多影像方法上持續改進，證實了過程層級稠密監督對多視角3D推理的有效性。

English

Multi-view 3D Visual Question Answering (MV3D-VQA) requires integrating partial observations into a coherent 3D scene representation and selecting informative viewpoints for multi-step spatial reasoning. However, current multimodal LLMs are typically trained with sparse, answer-level supervision, which often yields inconsistent cross-view reasoning and brittle view selection. We present DR-MV3D (Dense Reward for MV3D-VQA), a map-grounded learning framework that provides dense, verifiable rewards to supervise the reasoning process. Our approach decomposes MV3D-VQA into (i) allocentric global map construction, (ii) question-conditioned view-trajectory planning, and (iii) egocentric grounding for answer prediction. To make intermediate steps learnable without manual annotations, we introduce two rewards: a global consistency reward that aligns the predicted map with geometry-consistent pseudo targets from frozen 3D vision foundation models (e.g., VGGT + SAM3), and a local trajectory reward that supervises ordered viewpoint selection. We optimize the full pipeline with trajectory-level policy optimization (GRPO). Experiments on MindCube, VSI-Bench, and BLINK (MV) show that DR-MV3D consistently improves over strong multi-image baselines, supporting the effectiveness of process-level dense supervision for multi-view 3D reasoning.