GPT-6-Astra's Embodied Capabilities Face Major Test
GPT-6-Astra具身能力遭大考！传统基准拿满分，真机频碰壁
This research paper evaluates GPT-6-Astra's embodied intelligence capabilities across 12 simulated benchmarks and 8 real-world robot platforms. While it achieves high scores on structured tasks like LIBERO-Long++ and SIMPLE, real-world execution reveals significant limitations, including safety issues, precision challenges, and difficulty adapting to dynamic environments. The study emphasizes the gap between simulated performance and real-world deployment, highlighting the need for improved execution systems and task understanding.
The study evaluates GPT-6-Astra's embodied intelligence capabilities across 12 simulated benchmarks and 8 real-world robot platforms, highlighting its high performance in structured tasks but significant limitations in real-world execution, including safety, precision, and environmental adaptability.
Source: Machine Heart — Robotics on WeChat · 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
Reported success rate
87.3%
View original evidence
对于更复杂的灵巧操作，GPT-6-Astra 在 DexJoCo 上取得 87.3% 成功率，在 RoboDojo 上达到 81.9%。
Open source S19Reported success rate
81.9%
View original evidence
这一调整带来的效果立竿见影，以 RoboDojo 基准为例，其上的成功率从标准预算下的 27.6% 提高到 10× 预算下的 81.9%。
Open source S24Reported success rate
27.6%
View original evidence
这一调整带来的效果立竿见影，以 RoboDojo 基准为例，其上的成功率从标准预算下的 27.6% 提高到 10× 预算下的 81.9%。
Open source S24
- paper: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
技术报告地址：https://www.sensenova.ai/has-gpt-6-astra-achieved-embodied-intelligence 有意思的是，同一个 GPT-6-Astra，在简单解除限制之后，在多项传统具身评测中交出了满分或接近满分的成绩：其中在长程组合操作评测基准 LIBERO-Long++ 上达到 100%，在考察操作鲁棒性和泛化能力的 LIBERO-Pro 上达到 99.3%，人形机器人基准 SIMPLE 上达到 98.8%，RoboDojo 上也可以达到 81.9%。
Open source S7
除了上文已经提到的 LIBERO-Long++、LIBERO-Pro 和 SIMPLE，在超长程机器人操作基准 LIBERO-Ultra-Long 上为 98.9%。
Open source S18
对于更复杂的灵巧操作，GPT-6-Astra 在 DexJoCo 上取得 87.3% 成功率，在 RoboDojo 上达到 81.9%。
Open source S19
这一调整带来的效果立竿见影，以 RoboDojo 基准为例，其上的成功率从标准预算下的 27.6% 提高到 10× 预算下的 81.9%。
Open source S24
Source：Machine Heart — Robotics on WeChat · mp.weixin.qq.com