China Merchants River Island AI Lab Releases 'Yuwu Lion-ER-35B-A3B' Embodied Reasoning Model
不靠堆砌数据，让机器人看懂真实物理世界 | 招商局狮子山人工智能实验室发布「驭物 Lion-ER-35B-A3B」具身推理模型
The China Merchants River Island AI Lab has released the embodied reasoning visual language model 'Yuwu Lion-ER-35B-A3B'. It achieved improvements across ten public benchmarks, including specific gains in spatial placement and trajectory understanding tasks.
Source: Robotics — Chinese web discovery · Read original article ↗
English translation unavailable. Choose the original article to read it.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
training samples
312,380
View original evidence
「驭物™ Lion-ER-35B-A3B」的监督训练集共包含 312380条样本 ，训练时按能力需求分配训练预算，而不是由数据源的题量决定权重。
Open source S17records in material pool
≈4,300,000
View original evidence
确认存在能力缺口后，再从 覆盖31个数据源、约430万条记录 的素材池中，检索推理逻辑相同、场景不同的新素材，经构题核验后加入训练，再通过新一轮评测验证效果。
Open source S21data sources covered
31
View original evidence
确认存在能力缺口后，再从 覆盖31个数据源、约430万条记录 的素材池中，检索推理逻辑相同、场景不同的新素材，经构题核验后加入训练，再通过新一轮评测验证效果。
Open source S21
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
「驭物™ Lion-ER-35B-A3B」的监督训练集共包含 312380条样本 ，训练时按能力需求分配训练预算，而不是由数据源的题量决定权重。
Open source S17
确认存在能力缺口后，再从 覆盖31个数据源、约430万条记录 的素材池中，检索推理逻辑相同、场景不同的新素材，经构题核验后加入训练，再通过新一轮评测验证效果。
Open source S21
Source：Robotics — Chinese web discovery · sohu.com