Spatial-SSRL：通过自监督强化学习增强空间理解
Spatial-SSRL: Enhancing Spatial Understanding via Self-Supervised Reinforcement Learning
把普通 RGB/RGB-D 图像自动转成可验证奖励的自监督 pretext 任务（打乱图像块重排、翻转识别、裁剪块修复、区域深度排序、相对 3D 位置预测），可直接用于 LVLM 的 RLVR 训练，绕开人工标注与专用工具，是提升视觉语言模型空间理解的可复用数据构造思路。
原文依据 Spatial-SSRL automatically formulates five pretext tasks that capture 2D and 3D spatial structure
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论空间理解仍然是大视觉语言模型（LVLM）的一项短板。现有的监督微调（SFT）以及近期基于可验证奖励的强化学习（RLVR）流程依赖昂贵的监督、专用工具或受限环境，限制了其规模。我们提出 Spatial-SSRL，一种自监督强化学习范式，它直接从普通 RGB 或 RGB-D 图像中导出可验证信号。Spatial-SSRL 自动构造五个捕捉 2D 与 3D 空间结构的预训练任务：打乱图像块重排、翻转图像块识别、裁剪图像块修复、区域深度排序以及相对 3D 位置预测。这些任务提供易于验证、且无需人工或 LVLM 标注的真值答案。在我们的任务上训练可显著提升空间推理能力，同时保持通用视觉能力。在图像与视频两种设定下的七个空间理解基准上，Spatial-SSRL 相比 Qwen2.5-VL 基线分别取得 4.63%（3B）与 3.89%（7B）的平均准确率提升。我们的结果表明，简单的内在监督使 RLVR 能够规模化，并为 LVLM 获得更强的空间智能提供了一条切实可行的路径。
英文原文摘要 1207 字符
Spatial understanding remains a weakness of Large Vision-Language Models (LVLMs). Existing supervised fine-tuning (SFT) and recent reinforcement learning with verifiable rewards (RLVR) pipelines depend on costly supervision, specialized tools, or constrained environments that limit scale. We introduce Spatial-SSRL, a self-supervised RL paradigm that derives verifiable signals directly from ordinary RGB or RGB-D images. Spatial-SSRL automatically formulates five pretext tasks that capture 2D and 3D spatial structure: shuffled patch reordering, flipped patch recognition, cropped patch inpainting, regional depth ordering, and relative 3D position prediction. These tasks provide ground-truth answers that are easy to verify and require no human or LVLM annotation. Training on our tasks substantially improves spatial reasoning while preserving general visual capabilities. On seven spatial understanding benchmarks in both image and video settings, Spatial-SSRL delivers average accuracy gains of 4.63% (3B) and 3.89% (7B) over the Qwen2.5-VL baselines. Our results show that simple, intrinsic supervision enables RLVR at scale and provides a practical route to stronger spatial intelligence in LVLMs.