Hy-Nomas: Towards Free-form Interactive World Generation
Hy-Nomas keeps scene history available while limiting action commands to the moments they are meant to affect.
The video world model uses an 8B diffusion backbone to generate variable-length segments under free-form text instructions and continuous camera control. A vision-language Prompt Enhancer turns user requests and interaction history into structured conditions for the next segment. The authors report stronger performance than existing world models in human evaluation, though the abstract gives no scores or comparison details. They release model weights, code, benchmarks and a lightweight adaptation workflow.
HF Daily Papers' note
The video world model uses an 8B diffusion backbone to generate variable-length segments under free-form text instructions and continuous camera control. A vision-language Prompt Enhancer turns user requests and interaction history into structured conditions for the next segment. The authors report stronger performance than existing world models in human evaluation, though the abstract gives no scores or comparison details. They release model weights, code, benchmarks and a lightweight adaptation workflow.
HF Daily Papers' note
score 5