List your company

Vision-language Compositional Reasoning on ARO Benchmark Visual Genome Flickr30k MS-COCO (test)

89.3VG AttributionCapPa
51.75661.50371.2580.997Jun 13, 2023

Evaluation results

MethodLinks
CapPaArch=L/142023.0689.38699.399
CapArch=B/162023.0688.986.699.199
ARO BestArch=-2023.0688736046
CapPaArch=B/162023.0685.786.799.298.8
Blind dec.Arch=-2023.0683.786.298.898.7
NegCLIPArch=B/322023.0671819186
CLIPArch=B/322023.0663.259.159.447.3
CLIPArch=B/162023.0662.758.757.949.5
CLIP* (16k)Arch=L/142023.0657.835.940.231.5
CLIP* (8k)Arch=B/162023.0655.439.843.732.8
CLIP* (16k)Arch=B/162023.0653.239.745.537