IC-World: In-Context Generation for Shared World Modeling

2025-12-01

AI Summary

This work addresses the challenge of geometric and motion inconsistency across views in multi-view video generation. To this end, we propose IC-World, the first framework to systematically explore video world models for shared world modeling. Methodologically, IC-World leverages the contextual generation capability of large-scale video foundation models: given multi-view static images as input, it concurrently synthesizes dynamic video sequences across all views. We further introduce a group-relative policy optimization–based reinforcement learning mechanism, coupled with a novel dual reward model—explicitly enforcing scene-level geometric consistency and object-level motion consistency—for end-to-end optimization. Experiments demonstrate that IC-World significantly outperforms state-of-the-art methods on both geometric and motion consistency metrics, enabling high-fidelity, cross-view coherent dynamic content generation.

Technology Category

• Computer Vision: Computational Photography, Image & Video Synthesis
• Machine Learning: Multi-instance/Multi-view Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Video-based world models have recently garnered increasing attention for their ability to synthesize diverse and dynamic visual environments. In this paper, we focus on shared world modeling, where a model generates multiple videos from a set of input images, each representing the same underlying world in different camera poses. We propose IC-World, a novel generation framework, enabling parallel generation for all input images via activating the inherent in-context generation capability of large video models. We further finetune IC-World via reinforcement learning, Group Relative Policy Optimization, together with two proposed novel reward models to enforce scene-level geometry consistency and object-level motion consistency among the set of generated videos. Extensive experiments demonstrate that IC-World substantially outperforms state-of-the-art methods in both geometry and motion consistency. To the best of our knowledge, this is the first work to systematically explore the shared world modeling problem with video-based world models.

Problem

• Generates multiple videos from shared input images
• Enforces geometry and motion consistency across videos
• Systematically explores shared world modeling with video models

Innovation

• Activates in-context generation in large video models
• Uses reinforcement learning with Group Relative Policy Optimization
• Employs novel reward models for geometry and motion consistency