Our new X account is live! Follow @wizwand_team for updates

Confidence Calibration in Retrieval-Augmented Generation on NQ k=5 OOD (test)

0.248ECE

NAACL

Updated 4d ago

Evaluation Results

Method	Links
NAACL 2026.01		0.248	0.75
NAACL 2026.01		0.265	0.674
Label-only SFT 2026.01		0.273	0.653
NAACL 2026.01		0.276	0.628
Noise-aware 2026.01		0.289	0.667
Noise-aware 2026.01		0.29	0.605
Noise-aware 2026.01		0.304	0.641
CoT 2026.01		0.313	0.696
Vanilla 2026.01		0.322	0.706
Ensemble 2026.01		0.325	0.694
Label-only SFT 2026.01		0.329	0.604
Ensemble 2026.01		0.334	0.693
Label-only SFT 2026.01		0.335	0.667
NAACL 2026.01		0.335	0.64
Ensemble 2026.01		0.351	0.59
CoT 2026.01		0.352	0.67
Vanilla 2026.01		0.371	0.645
CoT 2026.01		0.373	0.621
Vanilla 2026.01		0.376	0.625
Noise-aware 2026.01		0.376	0.477
Ensemble 2026.01		0.398	0.597
Label-only SFT 2026.01		0.402	0.686
Vanilla 2026.01		0.435	0.608
CoT 2026.01		0.454	0.627