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

Object Hallucination Evaluation on POPE GQA (test)

84.72Average Accuracy

LLaVA-1.5 + HIRE

70.035273.847677.6681.4724Mar 11, 2026Mar 14, 2026Mar 17, 2026Mar 21, 2026Mar 24, 2026Mar 27, 2026Mar 31, 2026
Updated 18d ago

Evaluation Results

MethodLinks
2026.03
84.7288.8789.2184.6785.2180.6382.0785.5
2026.03
83.787.486.2183.3382.5480.378082.92
2026.03
81.54------83.38
2026.03
81.24------81.44
2026.03
80.89------82.37
2026.03
80.44------82.17
2026.03
80.28------81.02
2026.03
80.2585.6385.3878.7379.7876.478.1581.1
2026.03
80.0884.884.1679.3779.647676.8980.23
2026.03
79.9584.884.2379.2379.6375.8376.9380.26
2026.03
79.8------81.12
2026.03
79.59------80.2
2026.03
79.36------80.92
2026.03
79.23------80.16
2026.03
78.89------79.96
2026.03
78.76------80.79
2026.03
78.59------79.7
2026.03
78.11------80.29
2026.03
78.01------80.34
2026.03
77.98------80.32
2026.03
77.56------79.18
2026.03
77.31------79.08
2026.03
77.11------78.98
2026.03
71.99------73.89
2026.03
71.4------73.37
2026.03
71.13------73.19
2026.03
71.12------73.2
2026.03
70.81------73.08
2026.03
70.6------72.98
2025.08
-86.587.8374.7379.4568.7775.73-
2025.08
-8687.3374.579.2968.175.15-
2025.08
-87.588.5876.1780.2870.0776.42-
2025.08
-87.988.9475.279.6869.1375.91-
2025.08
-87.5388.6675.2779.7668.8776.43-
2025.08
-87.0788.2474.3779.169.275.9-
2025.08
-88.5789.3277.981.2272.5777.7-
2025.08
-86.5687.280.2380.6778.4379.27-
2025.08
-84.8784.0877.67875.776.38-
2025.08
-87.0386.3879.6780.1877.9778.88-
2025.08
-87.0786.9878.980.3776.7778.81-
2025.08
-86.185.479.880.1178.1378.81-
2025.08
-86.986.1579.638078.0378.76-
2025.08
-86.9386.3180.1380.3378.5779.1-
2025.08
-86.9387.3476.4779.371.676.04-
2025.08
-81.4381.673.7375.8171.9374.8-
2025.08
-8787.4775.0378.4272.1376.5-
2025.08
-87.4787.6575.9378.77376.71-
2025.08
-86.786.0476.279.1571.3775.93-
2025.08
-87.386.9975.5777.6674.0376.68-
2025.08
-87.5787.6976.5379.0673.8377.2-