List your company

Compositional Image Retrieval on GeneCIS (test)

22.7Focus Attribute R@1OSrCIR
9.713.07516.4519.825Oct 13, 2023Mar 1, 2024Jul 19, 2024Dec 6, 2024Apr 25, 2025Sep 12, 2025Jan 31, 2026

Evaluation results

MethodLinks
OSrCIRBackbone=ViT-G/14, Training-free=true, MLLM=GPT-4o2024.1222.736.44717.930.84216.928.436.72133.444.219.6
OSrCIRBackbone=OpenCLIP ViT-G/14, Model Configuration=w/ GPT4o2026.0122.736.44717.930.84216.928.436.72133.444.2—
ParacosmBackbone=OpenCLIP ViT-G/142026.0121.434.545.816.926.836.318.229.337.520.832.241.5—
OSrCIRBackbone=ViT-L/14, Training-free=true, MLLM=GPT-4o2024.1220.933.144.517.228.537.91523.634.218.430.638.317.9
CIReVL*Backbone=ViT-G/14, Training-free=true, MLLM=GPT-4o (two-stage)2024.1220.934.444.916.52939.815.125.633.418.531.641.417.8
CIReVLBackbone=ViT-G/14, Training-free=true2024.1220.53444.516.128.639.414.725.23318.131.24117.4
CIReVLBackbone=OpenCLIP ViT-G/142026.0120.53444.516.128.639.414.725.23318.131.241—
CIReVLBackbone=ViT-G/14*2023.1020.53444.516.128.639.414.725.23318.131.24117.4
CIReVL*Backbone=ViT-L/14, Training-free=true, MLLM=GPT-4o (two-stage)2024.1219.732.142.314.826.235.412.522.130.717.329.137.916.1
CIReVLBackbone=ViT-L/14, Training-free=true2024.1219.531.84214.42635.212.321.830.517.228.937.615.9
CIReVLBackbone=ViT-L/142023.1019.531.84214.42635.212.321.830.517.228.937.615.9
OSrCIRBackbone=ViT-B/32, Training-free=true, MLLM=GPT-4o2024.1219.432.742.816.427.738.115.725.735.818.230.139.417.4
LinCIRBackbone=ViT-G/14, Training-free=false2024.1219.13342.317.630.238.110.119.128.17.916.325.713.7
OSrCIRBackbone=OpenCLIP ViT-G/14, Model Configuration=w/ GPT4o, Evaluation Protocol=reproduced2026.0119.130.441.313.223.332.913.523.932.712.824.834.5—
CIReVLBackbone=RN50x42023.1019.131.341.514.726.33713.223.432.716.328.737.915.8
Combiner (CC3M)Backbone=RN50x42023.10193141.516.627.536.514.725.936.116.829.139.716.8
SEARLEBackbone=ViT-B/32, Training-free=false2024.1218.930.641.21323.833.712.22333.313.623.833.314.4
SEARLEBackbone=ViT-B/322023.1018.930.641.21323.833.712.22333.313.623.833.314.4
CIReVL*Backbone=ViT-B/32, Training-free=true, MLLM=GPT-4o (two-stage)2024.1218.229.740.715.126.136.114.924.533.616.428.137.916.2
CIReVLBackbone=ViT-B/32, Training-free=true2024.1217.929.440.414.825.835.814.624.333.316.127.837.615.9
CIReVLBackbone=ViT-B/322023.1017.929.440.414.825.835.814.624.333.316.127.837.615.9
Image OnlyBackbone=RN50x42023.1017.730.941.911.920.828.89.318.226.27.216.724.911.5
Context-I2WBackbone=ViT-L/14, Training-free=false2024.1217.230.541.716.428.337.18.717.926.97.71625.412.7
SEARLEBackbone=ViT-L/14, Training-free=false2024.1217.129.640.716.325.234.21222.230.91224.133.914.4
SEARLEBackbone=ViT-L/142023.1017.129.640.716.325.234.21222.230.91224.133.914.4
LinCIRBackbone=ViT-L/14, Training-free=false2024.1216.93041.516.22836.88.317.426.27.415.72512.2
SEARLEBackbone=OpenCLIP ViT-G/142026.0116.329.440.716.227.335.510.818.227.98.315.625.8—
Image + TextBackbone=RN50x42023.1015.626.337.112.622.93210.82131.211.321.530.312.6
Combiner (CIRR)Backbone=RN50x42023.1015.127.739.812.122.831.813.525.436.715.42839.614
Pic2WordBackbone=OpenCLIP ViT-G/142026.0112.523.433.711.721.930.99.919.327.48.618.226.1—
Text OnlyBackbone=RN50x42023.1010.220.529.69.517.626.46.516.822.46.213.921.48.1