Wan-R1: Verifiable-Reinforcement Learning for Video Reasoning

TL;DR

This paper introduces verifiable reward functions for reinforcement learning in video reasoning tasks, improving generalization and training stability in maze-solving and robotic navigation.

Contribution

It systematically studies reward design in RL for video reasoning, proposing verifiable rewards that enhance robustness and generalization over multimodal reward models.

Findings

Verifiable rewards improve exact match accuracy by 29.1% in maze tasks.

Verifiable rewards enhance trap-avoidance performance by 51.4%.

Multimodal reward models can cause degenerate solutions, unlike verifiable rewards.

Abstract

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.