Saved in:
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2602.09022 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912891520679936 |
|---|---|
| author | Wang, Zehan Wang, Tengfei Zhang, Haiyu Zuo, Xuhui Wu, Junta Wang, Haoyuan Sun, Wenqiang Wang, Zhenwei Cao, Chenjie Zhao, Hengshuang Guo, Chunchao Zhao, Zhou |
| author_facet | Wang, Zehan Wang, Tengfei Zhang, Haiyu Zuo, Xuhui Wu, Junta Wang, Haoyuan Sun, Wenqiang Wang, Zhenwei Cao, Chenjie Zhao, Hengshuang Guo, Chunchao Zhao, Zhou |
| contents | This work presents WorldCompass, a novel Reinforcement Learning (RL) post-training framework for the long-horizon, interactive video-based world models, enabling them to explore the world more accurately and consistently based on interaction signals. To effectively "steer" the world model's exploration, we introduce three core innovations tailored to the autoregressive video generation paradigm: 1) Clip-level rollout Strategy: We generate and evaluate multiple samples at a single target clip, which significantly boosts rollout efficiency and provides fine-grained reward signals. 2) Complementary Reward Functions: We design reward functions for both interaction-following accuracy and visual quality, which provide direct supervision and effectively suppress reward-hacking behaviors. 3) Efficient RL Algorithm: We employ the negative-aware fine-tuning strategy coupled with various efficiency optimizations to efficiently and effectively enhance model capacity. Evaluations on the SoTA open-source world model, WorldPlay, demonstrate that WorldCompass significantly improves interaction accuracy and visual fidelity across various scenarios. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_09022 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | WorldCompass: Reinforcement Learning for Long-Horizon World Models Wang, Zehan Wang, Tengfei Zhang, Haiyu Zuo, Xuhui Wu, Junta Wang, Haoyuan Sun, Wenqiang Wang, Zhenwei Cao, Chenjie Zhao, Hengshuang Guo, Chunchao Zhao, Zhou Computer Vision and Pattern Recognition This work presents WorldCompass, a novel Reinforcement Learning (RL) post-training framework for the long-horizon, interactive video-based world models, enabling them to explore the world more accurately and consistently based on interaction signals. To effectively "steer" the world model's exploration, we introduce three core innovations tailored to the autoregressive video generation paradigm: 1) Clip-level rollout Strategy: We generate and evaluate multiple samples at a single target clip, which significantly boosts rollout efficiency and provides fine-grained reward signals. 2) Complementary Reward Functions: We design reward functions for both interaction-following accuracy and visual quality, which provide direct supervision and effectively suppress reward-hacking behaviors. 3) Efficient RL Algorithm: We employ the negative-aware fine-tuning strategy coupled with various efficiency optimizations to efficiently and effectively enhance model capacity. Evaluations on the SoTA open-source world model, WorldPlay, demonstrate that WorldCompass significantly improves interaction accuracy and visual fidelity across various scenarios. |
| title | WorldCompass: Reinforcement Learning for Long-Horizon World Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.09022 |