E0.5: From Reconstructive to Future-Predictive 3D Foundations for Robotic World-
Overview
E0.5 is a geometry-native world-action model built on 3D foundation models, predicting future 3D states and actions. It achieves 89.90% success in clean settings and 50.09% in randomized settings without embodied-data pretraining. Further pretraining on 160k robot trajectories improves real-world manipulation success by 15 and 7 percentage points on AgileX Piper and Zerith H1.
From arXiv Robotics — research abstracts
Event evidence and corrections
1 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Reported quantity · Reported success rate: 85 percent · Basis not reported · Differing source assertions
“E0.5，在 AgileX Piper 双臂机器人和 Zerith H1 人形机器人上，经过机器人数据预训练后，它的整体平均成功率分别达到 85.0% 和 88.3% ，较 π0.5 分别高出 13.3 和 26.7 个百分点 —— 在与实力不俗的主流基线模型正面较量中，E0.5 依然展现出明显优势。”
Exact source · revision 1jiqizhixin
Reported quantity · Reported success rate: 88.3 percent · Basis not reported · Differing source assertions
“E0.5，在 AgileX Piper 双臂机器人和 Zerith H1 人形机器人上，经过机器人数据预训练后，它的整体平均成功率分别达到 85.0% 和 88.3% ，较 π0.5 分别高出 13.3 和 26.7 个百分点 —— 在与实力不俗的主流基线模型正面较量中，E0.5 依然展现出明显优势。”
Exact source · revision 1jiqizhixin
Reported quantity · Reported success rate: 96.7 percent · Basis not reported · Differing source assertions
“在 AgileX Piper 双臂机器人和 Zerith H1 人形机器人上，E0.5 的分布内任务 平均成功率分别达 到 96.7% 和 100.0% ；综合分布内与分布外设置，整体 平均成功率分别为 85.0% 和 88.3% ，较 π0.5 分别 提升 13.3 和 26.7 个百分点 。”
Exact source · revision 1jiqizhixin
Reported quantity · Reported success rate: 100 percent · Basis not reported · Differing source assertions
“在 AgileX Piper 双臂机器人和 Zerith H1 人形机器人上，E0.5 的分布内任务 平均成功率分别达 到 96.7% 和 100.0% ；综合分布内与分布外设置，整体 平均成功率分别为 85.0% 和 88.3% ，较 π0.5 分别 提升 13.3 和 26.7 个百分点 。”
Exact source · revision 1jiqizhixin
Reported quantity · Reported success rate: 85 percent · Basis not reported · Differing source assertions
“在 AgileX Piper 双臂机器人和 Zerith H1 人形机器人上，E0.5 的分布内任务 平均成功率分别达 到 96.7% 和 100.0% ；综合分布内与分布外设置，整体 平均成功率分别为 85.0% 和 88.3% ，较 π0.5 分别 提升 13.3 和 26.7 个百分点 。”
Exact source · revision 1jiqizhixin
Reported quantity · Reported success rate: 88.3 percent · Basis not reported · Differing source assertions
“在 AgileX Piper 双臂机器人和 Zerith H1 人形机器人上，E0.5 的分布内任务 平均成功率分别达 到 96.7% 和 100.0% ；综合分布内与分布外设置，整体 平均成功率分别为 85.0% 和 88.3% ，较 π0.5 分别 提升 13.3 和 26.7 个百分点 。”
Exact source · revision 1jiqizhixin
Reported quantity · Reported success rate: 86.7 percent · Basis not reported · Differing source assertions
“核心结果（不经大规模机器人数据预训练，直接进行任务后训练，对比 π0.5）： Piper 分布内 平均成功率达到 86.7%，H1 达到 93.3% ，相比同样未经具身预训练的 π0.5（46.7% 和 70.0%），分别 提升 40.0 和 23.3 个百分点 ； 充电器收纳在两个平台上分别 提升 20 和 40 个百分点 ——Piper 从 50% 提升至 70%，H1 从 40% 提升至 80%。”
Exact source · revision 1jiqizhixin
Evidence dependency changes (2)
- 2026-10-30T17:34:59.683Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T17:19:21.943Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
Report timeline
Follow attributed reports and material updates.
- Machine Heart — Robotics on WeChatSignalX-Era AI Releases Open-Source Native 4D World Action Model E0.5, Significantly Enhancing Robot Physical Interaction Capabilities
X-Era AI's E0.5 model, when tested on AgileX Piper dual-arm robots and Zerith H1 humanoid robots, significantly improved the success rate of complex physical interaction tasks through 3D geometric prediction and physical modeling. The model performed well in both in-distribution and out-of-distribution tasks, particularly demonstrating stronger generalization capabilities and operational stability when handling interactions with rigid and flexible objects.
- arXiv Robotics — research abstractsSignalE0.5: From Reconstructive to Future-Predictive 3D Foundations for Robotic World-Action Modeling
E0.5 is a geometry-native world-action model built on 3D foundation models, predicting future 3D states and actions. It achieves 89.90% success in clean settings and 50.09% in randomized settings without embodied-data pretraining. Further pretraining on 160k robot trajectories improves real-world manipulation success by 15 and 7 percentage points on AgileX Piper and Zerith H1.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).