List your company
ResearchFormerly WizWand

Interaction Recognition on ImageNet-1k 4 objects

91Interaction Recognition AccuracyMETAGAME
22.3640.185875.82May 7, 2026
Updated 5mo ago

Evaluation results

MethodLinks
METAGAMEModel (Size)=SigLIP-2 (ViT-B/32), Base Method=MaskSigLIP2026.0591
METAGAMEModel (Size)=SigLIP-2 (ViT-B/32), Base Method=Grad-ESigLIP2026.0591
METAGAMEModel (Size)=CLIP (ViT-B/16), Explanation Basis=Attention2026.0590
METAGAMEModel (Size)=CLIP (ViT-B/16), Explanation Basis=MaskCLIP2026.0589
FIxLIPModel (Size)=SigLIP-2 (ViT-B/32)2026.0588
METAGAMEModel (Size)=SigLIP-2 (ViT-L/16), Base Method=Grad-ESigLIP2026.0587
METAGAMEModel (Size)=MetaCLIP-2 (ViT-H/14), Explanation Basis=Attention2026.0586
METAGAMEModel (Size)=SigLIP-2 (ViT-L/16), Base Method=MaskSigLIP2026.0586
METAGAMEModel (Size)=CLIP (ViT-B/16), Explanation Basis=Grad-ECLIP2026.0584
METAGAMEModel (Size)=SigLIP-2 (ViT-B/32), Base Method=Attention2026.0582
METAGAMEModel (Size)=MetaCLIP-2 (ViT-H/14), Explanation Basis=Grad-ECLIP2026.0580
METAGAMEModel (Size)=SigLIP-2 (ViT-L/16), Base Method=Attention2026.0580
FIxLIPModel (Size)=SigLIP-2 (ViT-L/16)2026.0580
FIxLIPModel (Size)=CLIP (ViT-B/16)2026.0579
FIxLIPModel (Size)=MetaCLIP-2 (ViT-H/14)2026.0575
MaskSigLIPModel (Size)=SigLIP-2 (ViT-L/16)2026.0531
MaskSigLIPModel (Size)=SigLIP-2 (ViT-B/32)2026.0529
AttentionModel (Size)=CLIP (ViT-B/16)2026.0525
MaskCLIPModel (Size)=CLIP (ViT-B/16)2026.0525
Grad-ECLIPModel (Size)=CLIP (ViT-B/16)2026.0525
AttentionModel (Size)=MetaCLIP-2 (ViT-H/14)2026.0525
Grad-ECLIPModel (Size)=MetaCLIP-2 (ViT-H/14)2026.0525
AttentionModel (Size)=SigLIP-2 (ViT-L/16)2026.0525
Grad-ESigLIPModel (Size)=SigLIP-2 (ViT-L/16)2026.0525
AttentionModel (Size)=SigLIP-2 (ViT-B/32)2026.0525
Grad-ESigLIPModel (Size)=SigLIP-2 (ViT-B/32)2026.0525