List your company

Medical Logical Reasoning (Rule 2: is_mel_or_bcc) on HAM 5000 samples (test)

100F1 Score (NS-CL)GroundTruth (Oracle)
70.56878.20985.8593.491May 5, 2026

Evaluation results

MethodLinks
GroundTruth (Oracle)Supervision (%)=Oracle/GT2026.05100100
GlobalVAESupervision (%)=75%2026.0591.290.3
SlotVAE (k=1)Supervision (%)=75%2026.0571.770.7