Hallucination Detection on VQA-RAD Open-Ended benchmark leaderboard · SOTA2 Research

Hallucination Detection on VQA-RAD Open-Ended

83.13AUCVIHD

Evaluation Results

Method	Links
VIHD2026.05		83.13	72.01
VASE2026.05		76.98	69.09
VIHD2026.05		73.13	55.84
CEBaG2026.03		72.9	47.2
VASE2026.05		71.45	48.02
CEBaG2026.03		70.4	63.1
V-Loop2026.01		70	56.47
AvgProb2026.03		69.1	47.4
VASE2026.03		68.4	40.2
V-Loop2026.01		68.2	57.68
RadFlag2026.01		67.06	55.91
V-Loop2026.01		66.32	66.51
RadFlag2026.03		65.8	40.1
CEBaG2026.03		65.8	39.1
SE2026.03		64.1	62.8
SE2026.03		63.8	40.4
AvgEnt2026.01		63.5	60.17
SE2026.01		63.33	63.03
RadFlag2026.01		62.91	62.87
AvgProb2026.03		62.9	38.6
MaxEnt2026.01		62.8	60.17
MaxProb2026.05		62.52	53.54
MaxEnt2026.05		62.49	53.71
AvgProb2026.05		62.01	53.68
SE2026.03		61.9	65.3
VASE2026.03		61.7	59.6
SE2026.01		61.26	53.89
MaxEnt2026.01		61.17	52.11
AvgEnt2026.05		61.15	53.56
CEBaG2026.03		61	64.8
VASE2026.03		60.8	63.3
VASE2026.03		60.1	40.3
RadFlag2026.03		59.9	58.4
AvgEnt2026.01		58.89	51.44
RadFlag2026.03		58.4	63.6
AvgEnt2026.05		56.28	39.73
MaxEnt2026.05		56.27	39.74
AvgProb2026.05		56.23	39.63
MaxProb2026.05		56.21	39.67
RadFlag2026.03		54.5	39.3
RadFlag2026.05		53.27	36.33
SE2026.03		53.2	37.2
SE2026.01		52.4	49.64
RadFlag2026.01		51.9	48.35
SE2026.05		50.89	51.77
AvgProb2026.01		50.52	47.85
MaxProb2026.01		50.37	47.69
SE2026.05		50.37	35.93
RadFlag2026.05		50	49.3
MaxEnt2026.01		49.63	47.13
AvgEnt2026.01		49.48	47.07
AvgProb2026.03		43.7	48.1
AvgProb2026.03		39.9	48
AvgProb2026.01		39.88	42.12
MaxProb2026.01		38.4	38.05
MaxProb2026.01		36.86	45.37
AvgProb2026.01		36.36	43.47