ZhiXiang Future has released HiDream-O1-World, an interactive world model that supports multimodal inputs like text and images. The model excels in generating temporally and physically consistent worlds, a significant advancement in the field. It achieved top rankings on the WBench benchmark, particularly in spatial navigation and physical consistency, demonstrating its capability to create stable, realistic, and interactive environments. AI
IMPACT Sets new SOTA on interactive world modeling benchmarks, potentially accelerating AI-driven simulation and interactive entertainment.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Fudan University
- HiDream-O1-World
- Meituan LongCat team
- University of Tromsø – The Arctic University of Norway
- WBench
- Yao Ting
- ZhiXiang Future
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
