List your company

Mask-quality reviewer alignment on Corrupted masks (n=96) (test)

79.2Grade-Level Accuracy (Prec)Gem. 3 Pro
68.38471.1927476.808Jun 15, 2026

Evaluation results

MethodLinks
Gem. 3 ProRater category=VLM2026.0679.287.570.889.491.57.494.82
Gem. 3 FlashRater category=VLM2026.0678.187.566.789.695.16.093.99
Hum. (Gi)Rater category=Human2026.06767560.488.689.64.354.35
Gem. 3.1 ProRater category=VLM2026.067488.563.589.295.37.554.58
Hum. (C)Rater category=Human2026.0672.985.464.688.794.16.33.97
Hum. (V)Rater category=Human2026.0668.884.459.486.293.56.093.26