Share your thoughts, 1 month free Claude Pro on usSee more

Fact Verification on SciFact (test)

84Accuracy

Design B

Updated 1mo ago

Evaluation Results

Method	Links
Design B 2026.06		84	50.5	2.3	48.4	1.3
Design A rescue 2026.06		84	66.7	2.1	63.8	0.7
Design A 2-sided 2026.06		84	66.7	2.1	63.8	0.7
Design B 2026.06		84	66.7	2.1	63.8	0.7
Design A rescue 2026.06		83.2	50	1.8	48.9	1.8
Design A 2-sided 2026.06		83.2	50	1.8	48.9	1.8
SIFT 2026.06		80.9	64.6	-	63.1	-
SIFT 2026.06		79.4	48.2	-	47.1	-
Design A rescue 2026.06		79.4	46.6	0.5	47.6	0.5
Design A 2-sided 2026.06		79.4	46.6	0.5	47.6	0.5
Design B 2026.06		79.4	46.6	0.5	47.6	0.5
Direct 2026.06		79.4	41.5	-	-	-
SIFT 2026.06		78.6	46.1	-	47.1	-
Direct 2026.06		77.9	38.7	-	32	-
Direct 2026.06		71.8	45.6	-	50.9	-
Design B 2026.06		70.2	63.1	5.1	56.9	-1.1
NLI-as-judge (DeBERTa) 2026.06		69.7	-	-	-	-
Direct 2026.06		67.9	30.9	-	27.9	-
Design A rescue 2026.06		64.1	59.6	1.6	58	0
Design A 2-sided 2026.06		64.1	59.6	1.6	58	0
SIFT 2026.06		58.8	58	-	58	-