Announcement
EmbodiedFirst Exam introduced to test coding agents on simulated robotics tasks
ManlingLi_ introduced the benchmark for coding agents working on simulated robotics tasks across multiple suites.
TLDR
ManlingLi_ introduced EmbodiedFirst Exam on Oct. 19 as a test for coding agents on simulated robotics tasks. The project lists 1,090 tasks across 31 suites and 44 robots, with different levels of simulator access. Posts labeled “RoboDynamics: Real-Time Tasks” and “RoboEngineer: AutoResearch on Sim2Sim” offer examples of the broader exam, not results showing real-world robot performance.
Combined views
8.7K
7 Sources, first seen ago
