H3-World: Turning Language Understanding into World Control
Paper • 2609.01560 • Published • 37
🎮 H3-World: Turning Language Understanding into World Control
H3-World is the first interactive world model built on MiniMax-H3. Given an initial frame and keyboard controls, it generates action-controlled video with coordinated character and camera motion.
H3-World converts keyboard states into one language instruction per future video latent and uses directed attention routing to bind each instruction to its corresponding latent interval. Using 8,000 gameplay clips from ABot-World-Explorer-500h, it learns 65.6M LoRA parameters, only 0.199% of the 33B MiniMax-H3 backbone.
Checkpoint
| File | Description |
|---|---|
step-10000.safetensors |
Released rank-32 H3-World LoRA checkpoint |
Character controls use W, A, S, and D. Camera controls use I, J,
K, and L, with F indicating fast camera motion.
Usage
This checkpoint is a LoRA delta for MiniMax-H3. It requires the MiniMax-H3 base weights and H3-World's directed-attention patch; loading the LoRA through an unmodified MiniMax-H3 pipeline will not reproduce the reported behavior.
Please use the official H3-World code repository for setup, inference, and training instructions.
License
The H3-World LoRA checkpoint is released under Apache 2.0. MiniMax-H3 is governed by its own license terms.
Citation
@misc{chen2026h3worldturninglanguageunderstanding,
title={H3-World: Turning Language Understanding into World Control},
author={Danze Chen and Zeqing Wang and Ziyue Lin and Xingyi Yang and Yeying Jin},
year={2026},
eprint={2609.01560},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2609.01560},
}