List your company

Unknown entity vs. Visual evidence failure Classification on PopVQA and iNaturalist Attribution Tree

98.4GemmaVIS
29.96847.73465.583.266Jul 6, 2026

Evaluation results

MethodLinks
VISFeature Category=Pre-gen. features, Head=Lin.2026.0798.499.790.799.397
ATTN+LAST8Feature Category=Pre-gen. features, Head=Lin.2026.0768.46762.672.367.6
LAST8Feature Category=Pre-gen. features, Head=Lin.2026.0767.265.559.971.165.9
ATTNFeature Category=Pre-gen. features, Head=Lin.2026.0766.865.565.868.466.6
LAST8Feature Category=Pre-gen. features, Head=Tr.2026.0766.664.961.270.165.7
EOPFeature Category=Pre-gen. features, Head=Lin.2026.076260.361.662.561.6
TOKPROBFeature Category=Post-gen. and UQ, Head=Lin.2026.0760.262.257.156.759.1
OUTSEQFeature Category=Post-gen. and UQ, Head=Tr.2026.0757.560.755.755.157.2
OUTSEQFeature Category=Post-gen. and UQ, Head=Lin.2026.0756.259.454.253.755.9
ORACLE-UQFeature Category=Post-gen. and UQ2026.0755.951.548.753.552.4
MSPFeature Category=Post-gen. and UQ2026.075548.8474147.9
LAST8Feature Category=Pre-gen. features, Head=Tr.2026.0753.450.857.252.753.5
LAST8Feature Category=Pre-gen. features, Head=Lin.2026.0752.149.855.751.452.3
OUTSEQFeature Category=Post-gen. and UQ, Head=Tr.2026.0751.849.954.450.651.7
TOKPROBFeature Category=Post-gen. and UQ, Head=Lin.2026.0750.548.75349.250.4
RandomFeature Category=Post-gen. and UQ2026.0749.749.85049.849.8
OUTSEQFeature Category=Post-gen. and UQ, Head=Lin.2026.0749.647.85248.349.4
ATTN+LAST8Feature Category=Pre-gen. features, Head=Lin.2026.0748.64852.447.249
EOPFeature Category=Pre-gen. features, Head=Lin.2026.0747.345.250.346.947.4
ATTNFeature Category=Pre-gen. features, Head=Lin.2026.0743.944.648.242.944.9
MSPFeature Category=Post-gen. and UQ2026.0737.336.4302933.2
ORACLE-UQFeature Category=Post-gen. and UQ2026.0737.336.830.431.634
VISFeature Category=Pre-gen. features, Head=Lin.2026.0733.932.631.235.333.2
RandomFeature Category=Post-gen. and UQ2026.0732.634.130.228.731.4