List your company

Vision-Language Compositional Reasoning on Winoground 1.0 (test)

89.5Text ScoreHuman
6.09227.74649.471.054Jan 20, 2024Jun 9, 2024Oct 29, 2024Mar 20, 2025Aug 9, 2025Dec 29, 2025May 20, 2026

Evaluation results

MethodLinks
HumanSource=taken from original paper2026.0589.588.585.5
VQAScoreMethod Category=LLM-based, Backbone=CLIP-FlanT5-XXL2026.0560.855.344.8
BEiTScoreMethod Category=Encoder-based, Backbone=L/142026.0555.543.334.5
BEiTScoreMethod Category=Encoder-based, Backbone=L/14, Stage=without Stage12026.054835.327
IAISMethod category=Embedding-based2024.0142.519.516
FLEURMethod Category=LLM-based2026.054045.328.8
EXPERTMethod Category=LLM-based2026.05404628.3
BEiTScoreMethod Category=Encoder-based, Backbone=B/322026.0539.528.518.3
CACRMethod category=Embedding-based2024.0139.317.814.3
VinVLMethod Category=Encoder-based, Source=taken from original paper2026.0537.817.814.5
CLIP-SMethod Category=Encoder-based, Backbone=ViT-B2026.0535.511.38.5
BEiTScoreMethod Category=Encoder-based, Backbone=B/32, Stage=without Stage12026.0533.322.514.3
FLAVAMethod Category=Encoder-based, Variant=ITM, Source=taken from original paper2026.0532.320.514.3
PAC-S++Method Category=Encoder-based, Backbone=ViT-B2026.0531.88.86.5
PAC-S++Method Category=Encoder-based, Backbone=ViT-L2026.0531.5159.8
LongCLIP-SMethod Category=Encoder-based, Backbone=ViT-B2026.053197.3
CLIPMethod category=Embedding-based2024.0130.810.58
CLIP-SMethod Category=Encoder-based, Backbone=ViT-L2026.0530.811.59.5
LongCLIP-SMethod Category=Encoder-based, Backbone=ViT-L2026.0529.811.59.3
PAC-SMethod Category=Encoder-based, Backbone=ViT-B2026.0527106.8
PAC-SMethod Category=Encoder-based, Backbone=ViT-L2026.0526117.8
Random ChanceSource=taken from original paper2026.05252516.7
SPECSMethod Category=Encoder-based2026.059.37.83