机器之心 (微信公众号)T2·78 pts
- HiPhys-RSI reached 48.2% full-task success on 100 hidden household tasks, about 4x the previous best
- The challenge doubled to 100 tasks and 7 scenes, with submissions rising from 18 to 57
- Evaluation spans high-level reasoning, cross-room navigation and dual-arm manipulation, a hard test of embodied capability