INDEPENDENT MARKET INTELLIGENCE / CHINA ROBOTICS
industrial

ACE-CaP-0 Paper Reports 92.1% Success Across Eight Real-World Robot Tasks

The authors report 221 successful trials out of 240 across eight real-world manipulation tasks, with mean execution times of 0.97 to 2.27 minutes.

Software & embodied AIApplications & deploymentsIndustrial robots
ROBOPALLAS · 2026-11-16
Illustration introducing the ACE-CaP-0 research approach.

The ACE-CaP-0 paper evaluates a code-as-policies method on eight real-world manipulation tasks, including alignment, multi-object handling and bimanual coordination. The authors report 221 successful trials out of 240, or 92.1% overall success.

Mean execution times varied by task from 0.97 to 2.27 minutes. The authors say the frozen programs use runtime perception and control without a human or coding agent in the loop.

← More reports