From the Euler team

Better Robot Data, Measured: the Euler Readiness and Curation Benchmark

We measured Euler directly, against ground truth, with no policy training in the loop. On real robot video with a controlled fault layer, the readiness gate catches every injected timeline and camera fault (AUC 0.90 across all fault types) while auto-accepting 96% of clean episodes, zero-shot annotation agrees with human success labels 89% of the time, curation drops every exact duplicate while spanning four robot domains, and hybrid search returns the right episode first 87% of the time. Every number, measured.