RISE：世界行动模型的自适应想象

摘要

世界动作模型（WAMs）通过将未来世界演化纳入动作生成来改进规划，但现有方法为每个场景分配固定的想象预算。我们提出RISE（通过选择性推演精炼想象，Refining Imagination through Selective Rollout），一个系统级的自适应想象框架，根据持续推演的预期规划收益做出顺序化的推演/停止（Roll/Stop）决策。在每一步中，潜在评估器（Latent Evaluator）估计当前前缀所揭示的风险以及若想象继续则规划可能改进的程度，同时推演门控（Rollout Gate）将该预期收益与额外计算成本进行权衡。由于真实的驾驶日志仅暴露一个已实现的未来，我们进一步构建CounterDrive——一个包含多样结果和风险水平的反事实数据集，以丰富未来动态并提供局部化的风险监督。每个保留样本均经过专家验证和标注，涵盖轨迹有效性、事件起始和因果类别，为安全关键的世界建模研究提供可复用资源。在NAVSIM和nuScenes上的实验表明，RISE在减少不必要推演的同时实现了最佳的整体规划性能，额外的迁移结果支持其跨WAM架构的即插即用通用性。

English

World Action Models (WAMs) improve planning by incorporating future world evolution into action generation, yet existing methods allocate a fixed imagination budget to every scene. We propose RISE (Refining Imagination through SElective Rollout), a system-level adaptive imagination framework that makes sequential Roll/Stop decisions according to the expected planning benefit of continued rollout. At each step, a Latent Evaluator estimates the risk revealed by the current prefix and how much planning could improve if imagination continues, while a Rollout Gate weighs this expected benefit against additional computation cost. Since factual driving logs expose only one realized future, we further construct CounterDrive, a counterfactual dataset with diverse outcomes and risk levels, to enrich future dynamics and provide localized risk supervision. Each retained sample undergoes expert verification and annotation of trajectory validity, incident onset, and causal category, providing a reusable resource for safety-critical world-modeling research. Experiments on NAVSIM and nuScenes show that RISE achieves the best overall planning performance while reducing unnecessary rollout, with additional transfer results supporting its plug-in generality across WAM architectures.