perspectiva aiqbenchCercetare AI

The paper applies reinforcement learning to embodied video generation, training a hierarchical reward model on a roughly 235K-video preference dataset and using a PPO-style method to post-train flow-based generators. Its claims of improved physical fidelity and task logic rest on comparisons against prior methods, so readers should check the benchmarks and metrics before treating the gains as general. The approach of multi-dimensional reward modeling for physical plausibility may matter for robotics and simulation work that depends on realistic generated rollouts.

Sursă care acoperăarXiv

ReWorld: Multi-Dimensional Reward Modeling for Embodied Video Generation

arXiv3 noiembrie 20261 min de cititArticolul și perspectiva sunt în engleză.

arXiv:2601.12428v2 Announce Type: replace Abstract: Current embodied video generation models primarily emphasize visual generative quality while overlooking physical fidelity, dynamic consistency, and task logic, especially for contact-rich manipulation tasks, which limits their applicability to downstream tasks. In this paper, we introduce ReWorld, a framework aimed to employ reinforcement learning to align the video generation with physical realism, task completion capability, embodiment plausibility and visual quality. Specifically, we first construct a large-scale (~235K) video preference dataset and employ it to train a hierarchical reward model designed to capture multi-dimensional reward consistent with human preferences. We further propose a practical alignment algorithm that post-trains flow-based video generators using this reward through a computationally efficient PPO-style algorithm. Comprehensive experiments and theoretical analysis demonstrate that ReWorld significantly improves the physical fidelity, logical coherence, embodiment and visual quality of generated rollouts, outperforming previous methods.

Citește articolul complet pe arXiv

Citește pe arXiv ↗Întreabă despre acest articol