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

Polling-based Object Probing Evaluation (POPE) on GQA POPE Adversarial

81.92Accuracy

ICT w/o image

74.723276.591678.4680.3284Nov 22, 2024
Updated 1mo ago

Evaluation Results

MethodLinks
2024.11
81.9282.11
2024.11
81.6582
2024.11
81.681.53
2024.11
81.582.27
2024.11
80.8380.6
2024.11
80.2480.64
2024.11
80.0379.9
2024.11
80.0180.75
2024.11
76.0978.78
2024.11
75.4674.33
2024.11
75.0876.06
2024.11
7578.71