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

Object Hallucination Probing on GQA POPE (Adversarial)

82.73Accuracy

MESA

Updated 8d ago

Evaluation Results

Method	Links
MESA 2026.04		82.73	-	-	-	-	-	-	-	-	-	85.94	78.27	81.93
ICT 2026.04		81.61	-	-	-	-	-	-	-	-	-	84.61	77.21	80.75
ICD 2026.04		81	-	-	-	-	-	-	-	-	-	84.24	76.27	80.06
VTI 2026.04		80.47	-	-	-	-	-	-	-	-	-	87.52	71.07	78.44
VAF 2026.04		80.03	-	-	-	-	-	-	-	-	-	83.3	76.53	79.75
VCD 2026.04		80.01	-	-	-	-	-	-	-	-	-	77.86	83.85	80.75
MESA 2026.04		79.6	-	-	-	-	-	-	-	-	-	74.92	89	81.35
ICT 2026.04		79.13	-	-	-	-	-	-	-	-	-	78.17	84.16	81.07
VAF 2026.04		79.1	-	-	-	-	-	-	-	-	-	75.66	85.8	80.41
VTI 2026.04		78.33	-	-	-	-	-	-	-	-	-	74.8	85.47	79.78
VCD 2026.01		77.3	-	-	-	-	-	-	-	-	-	78.98	74.4	76.62
VCD 2026.04		76.09	-	-	-	-	-	-	-	-	-	70.83	88.75	78.78
ICD 2026.04		75.9	-	-	-	-	-	-	-	-	-	73.84	80.2	76.89
Regular Decoding 2026.01		75.47	-	-	-	-	-	-	-	-	-	78.13	70.73	74.25
SDCD 2026.01		74.47	-	-	-	-	-	-	-	-	-	75.45	72.53	73.96
Nullu 2026.04		73.61	-	-	-	-	-	-	-	-	-	94.13	50.33	65.59
Vanilla 2026.04		73.53	-	-	-	-	-	-	-	-	-	93.9	50.33	65.53
Nullu 2026.04		72.1	-	-	-	-	-	-	-	-	-	66.22	90.2	76.37
Vanilla 2026.04		67.4	-	-	-	-	-	-	-	-	-	61.3	94.33	74.31
Baseline 2025.06		-	81.17	63.17	63.3	64.1	67.9	62.1	-	-	64.11	-	-	-
VCD 2025.06		-	-	67.33	66.97	63.87	67.83	63.37	-	-	65.87	-	-	-
CLAIM 2025.06		-	77.57	75.93	76.27	75.23	76.47	-	-	-	76.29	-	-	-
Baseline 2025.06		-	82.63	78.97	69.73	73.07	-	-	61.73	71.63	71.52	-	-	-
VCD 2025.06		-	-	79.57	73.47	71.83	-	-	54.97	76.17	71.2	-	-	-
CLAIM 2025.06		-	79.97	76.77	79.53	62.47	-	-	77.63	-	75.27	-	-	-