立体世界模型：相机引导的立体视频生成
Stereo World Model: Camera-Guided Stereo Video Generation
提出相机帧RoPE和立体感知注意力分解，在RGB模态下直接生成具有视差一致性的立体视频，可复用于需要相机控制的多视角视频生成任务，并支持长视频蒸馏。
原文依据 a unified camera-frame RoPE that augments latent tokens with camera-aware rotary positional encoding
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论我们提出 StereoWorld，一种相机条件化的立体世界模型，联合学习外观与双目几何，实现端到端的立体视频生成。与单目 RGB 或 RGBD 方法不同，StereoWorld 完全在 RGB 模态内运行，同时直接从视差中定位几何。为高效实现一致的立体生成，我们的方法引入两项关键设计：(1) 统一的相机帧 RoPE，用相机感知的旋转位置编码增强潜变量 token，实现相对、视角与时间一致的调节，并通过稳定的注意力初始化保留预训练视频先验；(2) 立体感知注意力分解，将完整的 4D 注意力分解为 3D 视图内注意力加水平行注意力，利用极线先验以显著更低的计算量捕获视差对齐的对应关系。在基准测试中，StereoWorld 在立体一致性、视差精度和相机运动保真度上优于强单目后转换流程，生成速度提升 3 倍以上，视角一致性额外提升 5%。除基准测试外，StereoWorld 无需深度估计或修复即可实现端到端双目 VR 渲染，通过度量尺度的深度定位增强具身策略学习，并兼容长视频蒸馏以扩展交互式立体合成。
英文原文摘要 1427 字符
We present StereoWorld, a camera-conditioned stereo world model that jointly learns appearance and binocular geometry for end-to-end stereo video generation.Unlike monocular RGB or RGBD approaches, StereoWorld operates exclusively within the RGB modality, while simultaneously grounding geometry directly from disparity. To efficiently achieve consistent stereo generation, our approach introduces two key designs: (1) a unified camera-frame RoPE that augments latent tokens with camera-aware rotary positional encoding, enabling relative, view- and time-consistent conditioning while preserving pretrained video priors via a stable attention initialization; and (2) a stereo-aware attention decomposition that factors full 4D attention into 3D intra-view attention plus horizontal row attention, leveraging the epipolar prior to capture disparity-aligned correspondences with substantially lower compute. Across benchmarks, StereoWorld improves stereo consistency, disparity accuracy, and camera-motion fidelity over strong monocular-then-convert pipelines, achieving more than 3x faster generation with an additional 5% gain in viewpoint consistency. Beyond benchmarks, StereoWorld enables end-to-end binocular VR rendering without depth estimation or inpainting, enhances embodied policy learning through metric-scale depth grounding, and is compatible with long-video distillation for extended interactive stereo synthesis.