List your company

Evaluation of metrics' alignment with human judgment on RadEvalX

0.927Spearman CorrelationAgentsEval
0.0631760.2874380.51170.735962Jan 23, 2026Feb 15, 2026Mar 11, 2026Apr 3, 2026Apr 27, 2026May 20, 2026Jun 13, 2026

Evaluation results

MethodLinks
AgentsEval2026.010.9271.867——
Agent(det)2026.010.9191.728——
Agent(sim)2026.010.91.181——
B-Score2026.010.7890.909——
ROUGE-12026.010.654.412——
QAScore2026.060.4612—0.45070.391
METEOR2026.010.4525.004——
GREEN2026.060.4022—0.41940.3148
CheXbert2026.060.3779—0.3870.2896
CHRF2026.010.3574.86——
RaTEScore2026.060.3551—0.35390.2651
BERTScore2026.060.2525—0.25750.1901
RadCliQ2026.060.2153—0.20170.1586
BLEU-22026.060.1742—0.18510.1265
BLEU2026.010.178.155——
RadGraph F12026.060.1692—0.17430.1238
BLEU-42026.060.0964—0.1190.0728