Wan-R1: Verifiable-Reinforcement Learning for Video Reasoning
For researchers working on video reasoning and RL-based video generation, this work establishes verifiable reward design as a critical enabler for robust performance, addressing a previously understudied bottleneck.
The paper tackles the challenge of improving spatial reasoning and multi-step planning in video generation models using reinforcement learning. By adapting GRPO to flow-based video models and designing verifiable reward functions, they achieve a 29.1% improvement in exact match accuracy on complex 3D mazes and 51.4% on trap-avoidance tasks over SFT baselines.
Video generation models produce visually coherent content but struggle with tasks requiring spatial reasoning and multi-step planning. Reinforcement learning (RL) offers a path to improve generalization, but its effectiveness in video reasoning hinges on reward design -- a challenge that has received little systematic study. We investigate this problem by adapting Group Relative Policy Optimization (GRPO) to flow-based video models and training them on maze-solving and robotic navigation tasks. We first show that multimodal reward models fail catastrophically in this setting. To address this, we design verifiable reward functions grounded in objective task metrics. For structured game environments, we introduce a multi-component trajectory reward. For robotic navigation, we propose an embedding-level verifiable reward. Our experiments show that RL fine-tuning with verifiable rewards improves generalization. For example, on complex 3D mazes, our model improves exact match accuracy by 29.1\% over the SFT baseline, and on trap-avoidance tasks by 51.4\%. Our systematic reward analysis reveals that verifiable rewards are critical for stable training, while multimodal reward models could lead to degenerate solutions. These findings establish verifiable reward design as a key enabler for robust video reasoning. Code will be publicly available.