GrndCtrl: 通过自监督奖励对齐实现世界模型的 grounding
GrndCtrl: Grounding World Models via Self-Supervised Reward Alignment
AI总结：
GrndCtrl通过自监督奖励对齐提升世界模型的几何 grounding，实现更稳定的空间一致性与导航性能。
摘要：
最近视频世界建模的进步使大规模生成模型能够以高视觉保真度模拟具身环境，为预测、规划和控制提供强有力的先验知识。然而，尽管这些模型具有现实感，但它们往往缺乏几何 grounding，限制了其在需要空间一致性和稳定性的导航任务中的应用。我们介绍了强化学习与世界 grounding（RLWG），一种自监督的后训练框架，通过几何和感知奖励将预训练的世界模型与可验证的结构对齐。类似于语言模型中的可验证反馈强化学习（RLVR），RLWG可以使用多种奖励来衡量姿态循环一致性、深度重投影和时间一致性。我们通过GrndCtrl实现了这一框架，这是一种基于组相对策略优化（GRPO）的奖励对齐适应方法，产生保持稳定轨迹、一致几何和可靠重放的具身导航世界模型。类似于大语言模型中的后训练对齐，GrndCtrl利用可验证的奖励来弥合生成预训练和 grounded 行为之间的差距，在户外环境中实现了优于监督微调的空间一致性和导航稳定性。
英文摘要：
Recent advances in video world modeling have enabled large-scale generative models to simulate embodied environments with high visual fidelity, providing strong priors for prediction, planning, and control. Yet, despite their realism, these models often lack geometric grounding, limiting their use in navigation tasks that require spatial coherence and stability. We introduce Reinforcement Learning with World Grounding (RLWG), a self-supervised post-training framework that aligns pretrained world models with a physically verifiable structure through geometric and perceptual rewards. Analogous to reinforcement learning from verifiable feedback (RLVR) in language models, RLWG can use multiple rewards that measure pose cycle-consistency, depth reprojection, and temporal coherence. We instantiate this framework with GrndCtrl, a reward-aligned adaptation method based on Group Relative Policy Optimization (GRPO), yielding world models that maintain stable trajectories, consistent geometry, and reliable rollouts for embodied navigation. Like post-training alignment in large language models, GrndCtrl leverages verifiable rewards to bridge generative pretraining and grounded behavior, achieving superior spatial coherence and navigation stability over supervised fine-tuning in outdoor environments.
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。
