RISE: Adaptive Imagination for World Action Models

arXiv:2608.20430v1 Announce Type: new Abstract: 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 (extbf{R}efining extbf{I}magination through extbf{SE}lective Rollout), a system-level adaptive imagination framework that makes sequential extsc{Roll}/extsc{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 extbf{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.

Related

• Light-WAM: Efficient World Action Models with State-Fusion Action Decoding
• ReWorld: Learning Better Representations for World Action Models
• SG-WAM: Self-Guided World Modeling in Geometry-Aware Policy Space
• Is the Future Compatible? Diagnosing Dynamic Consistency in World Action Models
• Robust-WAM: Bridging Generative Pretraining and Semantic Foresight in World-Action Models

Source: arXiv cs.CV | 2026-08-24