Table 5.

Human raters show good to excellent agreement on the held-out test set. Determining agreement on the same test set the AI model is evaluated on can help provide a better baseline for expected performance.

GRS Domain	ICC (2,3)	SEM (2,3)	ICC (2,1)	SEM (2,1)
Respect for Tissue	0.78	0.44	0.54	0.63
Time and Motion	0.81	0.41	0.58	0.61
Quality of Final Product	0.93	0.30	0.82	0.49
Overall Performance	0.86	0.30	0.68	0.30