ReaForest: Fostering Generative Video Reasoning for Spatial Planning

Verbal logic and visual mental simulation are two essential components of human intelligence. Modern Large Language Models (LLMs) have demonstrated strong verbal reasoning capabilities through textual Chain-of-Thought (CoT) reasoning. In contrast, current Video Generation Models (VGMs) struggle with visual reasoning tasks such as spatial planning. We attribute this limitation to two fundamental gaps: (i) VGMs are predominantly trained on general-purpose video corpora emphasizing perceptual fidelity over visual reasoning, leaving reasoning abilities underdeveloped; (ii) most VGMs generate videos in a single pass without mechanisms to explore alternative reasoning trajectories and to revise intermediate errors. Motivated by these limitations, we introduce ReaForest, a framework that fosters the reasoning capacity of VGMs in spatial planning through both training-time activation and inference-time scaling. ReaForest comprises three key components: (1) ReaGen-27k, a dataset covering diverse spatial planning tasks that require multi-step reasoning, which activates basic reasoning capabilities of VGMs for spatial planning; (2) Reflective Entropy-Aware Test-Time Scaling (ReaTTS), an inference framework that evolves multiple reasoning branches while enabling failure recovery; (3) Hierarchical constraint verification, which provides actionable feedback for ReaTTS based on decomposed constraints. Extensive experiments demonstrate that ReaForest substantially surpasses advanced textual reasoning models (e.g., Gemini-2.5-Pro) and video generation models (e.g., Sora-2). ReaForest exhibits emergent properties including self-correction, parallel thinking, and scalable reasoning, advancing VGMs toward human-like visual mental simulation.

Modern AI systems can reason well with text, but they still struggle with visual planning tasks that humans solve naturally, such as navigating mazes or planning movements in complex environments. In this work, we introduce ReaForest, a framework that helps video generation models “think visually” by reasoning through sequences of generated video frames. ReaForest combines three key ideas: a large-scale training dataset for spatial planning, a test-time reasoning strategy that explores multiple possible solution paths while correcting mistakes, and a verification system that checks whether each step follows task constraints. Together, these components allow video models to progressively plan and revise actions rather than generating videos in a single pass. Experiments on multiple spatial reasoning benchmarks show that ReaForest significantly outperforms existing video generation models and even strong multimodal language models on tasks such as maze navigation and Sokoban puzzles. Our findings suggest that video generation models can evolve from content creation tools into visual reasoning systems capable of human-like mental simulation.