arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.14225cs.AI
收藏

文本优先于视觉:针对超高清遥感理解的代理强化学习在超高清遥感理解中的知识注入至关重要

Text Before Vision: Staged Knowledge Injection Matters for Agentic RLVR in Ultra-High-Resolution Remote Sensing Understanding

National University of Defense Technology, China(国防科技大学);Beijing University of Posts and Telecommunications, China(北京邮电大学);University of the Chinese Academy of Sciences, China(中国科学院大学);Sichuan University, China(四川大学);Wuhan University, China(武汉大学);Chinese Academy of Science, China(中国科学院);Tsinghua University, China(清华大学);Shanghai Artificial Intelligence Laboratory, China(上海人工智能实验室);Renmin University of China, China(中国人民大学)
Fengxiang Wang, Mingshuo Chen, Yueying Li, Yajie Yang, Yuhao Zhou, Di Wang, Yifan Zhang, Haoyu Wang, Haiyan Zhao, Hongda Sun, Long Lan, Jun Song, Yulin Wang, Ji… 展开作者

Fengxiang Wang, Mingshuo Chen, Yueying Li, Yajie Yang, Yuhao Zhou, Di Wang, Yifan Zhang, Haoyu Wang, Haiyan Zhao, Hongda Sun, Long Lan, Jun Song, Yulin Wang, Jing Zhang, Wenlong Zhang, Bo Du

AI总结:

本文提出分阶段知识注入方法,利用文本引导提升超高清遥感理解的视觉推理能力,实现XLRS-Bench上的高准确率。

摘要:

超高清(UHR)遥感(RS)的多模态推理通常受到视觉证据获取的瓶颈:模型需要在庞大的像素空间中定位微小的任务相关区域。虽然使用放大工具的代理强化学习与可验证奖励(RLVR)提供了一条前进的路径,但我们发现标准强化学习在没有结构化领域先验知识的情况下难以在这些庞大的视觉空间中导航。在本文中,我们研究了后训练范式之间的相互作用:在UHR RS基准上比较冷启动监督微调(SFT)、RLVR和代理RLVR。我们的受控研究得出一个反直觉的发现:高质量的纯文本地球科学问答是UHR视觉推理增益的主要驱动因素。尽管没有图像,领域特定的文本注入了必要的概念、机理解释和决策规则,以指导视觉证据检索。基于此,我们提出了一种分阶段的知识注入配方:(1)使用可扩展的、基于知识图谱验证的地球科学文本问答进行冷启动,以建立推理结构;(2)在相同的困难UHR图像-文本示例上进行SFT预热,以稳定并放大后续基于工具的RL。这种方法在XLRS-Bench上实现了60.40%的Pass@1,显著优于更大的通用模型(例如,GPT-5.2、Gemini 3.0 Pro、Intern-S1),并建立了新的最先进的水平。

英文摘要:

Multimodal reasoning for ultra-high-resolution (UHR) remote sensing (RS) is usually bottlenecked by visual evidence acquisition: the model necessitates localizing tiny task-relevant regions in massive pixel spaces. While Agentic Reinforcement Learning with Verifiable Rewards (RLVR) using zoom-in tools offers a path forward, we find that standard reinforcement learning struggles to navigate these vast visual spaces without structured domain priors. In this paper, we investigate the interplay between post-training paradigms: comparing Cold-start Supervised Fine-Tuning (SFT), RLVR, and Agentic RLVR on the UHR RS benchmark.Our controlled studies yield a counter-intuitive finding: high-quality Earth-science text-only QA is a primary driver of UHR visual reasoning gains. Despite lacking images, domain-specific text injects the concepts, mechanistic explanations, and decision rules necessary to guide visual evidence retrieval.Based on this, we propose a staged knowledge injection recipe: (1) cold-starting with scalable, knowledge-graph-verified Earth-science text QA to instill reasoning structures;and (2) "pre-warming" on the same hard UHR image-text examples during SFT to stabilize and amplify subsequent tool-based RL. This approach achieves a 60.40% Pass@1 on XLRS-Bench, significantly outperforming larger general purpose models (e.g., GPT-5.2, Gemini 3.0 Pro, Intern-S1) and establishing a new state-of-the-art.

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。

↑