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

Hallucination Evaluation on POPE MSCOCO, A-OKVQA, GQA average (Adversarial)

84.11Accuracy

Vanilla

76.944478.804780.66582.5253Feb 14, 2026
Updated 3mo ago

Evaluation Results

MethodLinks
2026.02
84.1180.2591.3585.24
2026.02
84.0980.3891.0485.17
2026.02
83.5578.9692.685
2026.02
83.0578.8591.2984.4
2026.02
82.3977.1793.1184.17
2026.02
80.1874.4793.8282.72
2026.02
80.1574.4693.882.7
2026.02
78.7172.0494.581.66
2026.02
78.6872.7192.5981.3
2026.02
78.4871.5595.1781.61
2026.02
78.4671.6194.9181.54
2026.02
77.8171.0195.8981.36
2026.02
77.4671.8793.2280.77
2026.02
77.2370.9595.1380.93
2026.02
77.2270.7193.5380.44