List your company

Question Answering on HotpotQA (Acc, AUROC, Brier, ECE)

63AccuracyRLVR
38.76845.05951.3557.641Jul 22, 2025

Evaluation results

MethodLinks
RLVR2025.07630.50.370.37
RLVR + BCE Classifier2025.07630.660.220.07
RLVR + Brier Classifier2025.07630.650.220.09
RLVR + Probe2025.07630.550.240.1
Answer Probability2025.07630.720.360.36
RLCR2025.0762.10.690.210.03
Base2025.0739.70.540.530.53