ROVER: 面向对象中心视觉证据的路由用于基于多图像推理
ROVER: Routing Object-Centric Visual Evidence for Grounded Multi-Image Reasoning
AI总结：
提出ROVER，一种轻量级可学习插件，通过对象中心差分注意力聚合上下文、蒸馏图像内线索并路由历史感知证据，实现高效全局视觉证据路由，在多图像推理中提升答案和定位精度。
摘要：
多模态大语言模型（MLLMs）越来越多地定位和交错视觉证据以进行审慎推理。基于定位的方法通常通过将裁剪的图像块或感兴趣区域（RoI）特定特征注入推理上下文来关注RoI。然而，这种设计可能削弱整体场景理解和对象间关系，同时导致解码成本随RoI数量和大小增加而增加。或者，自适应视觉特征选择通常需要细粒度监督或复杂启发式方法。为解决这些限制，我们提出ROVER（面向对象中心视觉证据的路由用于基于多图像推理），一种轻量级、可学习的插件，用于高效的全局视觉证据路由。在每次对象定位预测时，ROVER注入一个步骤特定的令牌三元组，以协同地：(i) 聚合正在进行的推理上下文，(ii) 通过对象中心差分注意力将图像内线索蒸馏到视觉工作空间中，以及(iii) 在该空间内跨对象和图像路由并整合历史感知证据以供后续推理。我们将ROVER集成到Qwen2.5-VL-7B中，并开发了一个交错的SFT到GRPO训练流程。严格遵循原始数据集和评估协议，我们的方法在MM-GCoT（+4.8%答案准确率，+14.6%定位准确率）和VideoEspresso（+8.6%答案准确率）上取得了最佳性能。在VideoEspresso上训练的模型表现出强大的迁移能力，在多个基准测试上平均比基础模型高出+4.7%。
英文摘要：
Multimodal Large Language Models (MLLMs) have increasingly localized and interleaved visual evidence for deliberative reasoning. Grounding-based approaches typically focus on regions of interest (RoIs) by injecting cropped image patches or RoI-specific features into the reasoning context. However, such designs can weaken holistic scene understanding and inter-object relations, while incurring decoding costs that scale with the number and size of RoIs. Alternatively, adaptive visual feature selection often requires fine-grained supervision or complex heuristics. To address these limitations, we propose ROVER (Routing Object-centric Visual Evidence for grounded multi-image Reasoning), a lightweight, learnable plugin for efficient global visual evidence routing. Upon each object grounding prediction, ROVER injects a step-specific token triplet to synergistically: (i) aggregate the ongoing reasoning context, (ii) distill intra-image cues into a visual working space via object-centric differential attention, and (iii) route and integrate history-aware evidence across objects and images within this space for subsequent reasoning. We integrate ROVER into Qwen2.5-VL-7B and develop an interleaved SFT-to-GRPO training pipeline. Strictly adhering to the original datasets and evaluation protocols, our method achieves the best performance on MM-GCoT (+4.8% answer accuracy, +14.6% grounding accuracy) and VideoEspresso (+8.6% answer accuracy). The VideoEspresso-trained model demonstrates strong transferability, outperforming the base model by +4.7% on average across diverse benchmarks.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
