Aerial-R1: Reinforcing Aerial Reasoning and Segmentation in MLLMs

查看原文 / View Original 编辑 / Edit

Computer Science > Computer Vision and Pattern Recognition arXiv:2610.33056 (cs) [Submitted on 19 Oct 2026] Title: Aerial-R1: Reinforcing Aerial Reasoning and Segmentation in MLLMs Authors: Jisheng Dang, Huaiyuan Peng, Yixuan Wang, Kun Wang, Zhang Li, Tat-Seng Chua.

计算机科学 > 计算机视觉与模式识别 arXiv:2610.33056 (cs) [提交于 2026 年 10 月 19 日] 标题：Aerial-R1：增强多模态大模型（MLLM）中的航空推理与分割能力作者：Jisheng Dang, Huaiyuan Peng, Yixuan Wang, Kun Wang, Zhang Li, Tat-Seng Chua。

Abstract: While Multimodal Large Language Models (MLLMs) have revolutionized general visual perception, existing methods suffer from severe hallucination in the aerial domain, where extreme scale variations and visual ambiguity abound. Existing Supervised Fine-Tuning (SFT) paradigms force models to mimic ground truth without modeling uncertainty, often resulting in blind guessing on tiny or unrecognizable objects in aerial scenes.

摘要：尽管多模态大模型（MLLM）彻底改变了通用视觉感知，但现有方法在航空领域存在严重的幻觉问题，该领域充斥着极端的尺度变化和视觉模糊性。现有的监督微调（SFT）范式强迫模型模仿真实标签而无需建模不确定性，这往往导致模型在面对航空场景中微小或无法识别的物体时进行盲目猜测。

In this paper, we present Aerial-R1, the first aerial MLLM framework to reinforce visual reasoning capabilities for aerial reasoning and segmentation. Drawing inspiration from the reasoning-incentivizing mechanisms of DeepSeek-R1, we find that trustworthy perception stems from consistency across views, establishing a visual parallel to the Chain-of-Thought in logic.

在本文中，我们提出了 Aerial-R1，这是首个旨在增强航空推理与分割视觉推理能力的航空 MLLM 框架。受 DeepSeek-R1 推理激励机制的启发，我们发现可信的感知源于跨视图的一致性，从而建立了逻辑中思维链（Chain-of-Thought）的视觉对应关系。

To effectively adapt policy optimization for non-autoregressive mask generation, we introduce Hybrid-View Group Advantage-Weighted Regression (H-GAWR). This strategy constructs dynamic groups of high-resolution and low-resolution views, seamlessly bridging group-relative advantage estimation with advantage-weighted policy extraction.

为了有效地将策略优化适配于非自回归掩码生成，我们引入了混合视图组优势加权回归（H-GAWR）。该策略构建了高分辨率和低分辨率视图的动态组，无缝连接了组相对优势估计与优势加权策略提取。

By enforcing consensus on clear inputs and strictly penalizing hallucinations on ambiguous ones, Aerial-R1 develops an explicitly incentivized refusal mechanism. Extensive experiments on VSAI-Ref, the first aerial MLLM benchmark dedicated to jointly evaluating dense reasoning and hallucination robustness, demonstrate that Aerial-R1 significantly outperforms SFT baselines in segmentation precision while substantially mitigating fatal hallucinations in low-information scenarios, paving the way for more reliable and trustworthy aerial agents.

通过在清晰输入上强制达成共识，并对模糊输入上的幻觉进行严格惩罚，Aerial-R1 开发了一种显式激励的拒绝机制。在 VSAI-Ref（首个专门用于联合评估密集推理和幻觉鲁棒性的航空 MLLM 基准）上进行的广泛实验表明，Aerial-R1 在分割精度上显著优于 SFT 基线，同时在低信息场景中大幅减轻了致命的幻觉，为更可靠、更可信的航空智能体铺平了道路。