List your company

Selective Answering on HotpotQA RAG v3

0.1642Risk@30SURE-RAG
0.1580360.1996430.241250.282857May 5, 2026

Evaluation results

MethodLinks
SURE-RAGCalibration=Standard2026.050.16420.360.5457
SURE-RAGCalibration=Calibrated2026.050.1670.3510.5493
DeBERTaPooling=mean-pool2026.050.25880.37830.535
DeBERTaPooling=max-pool2026.050.31830.40540.5344