List your company

Image-Text Retrieval on MSCOCO (IR@1, TR@1)

81.9IR@1MDPD
18.77235.16151.5567.939Mar 9, 2026Mar 15, 2026Mar 21, 2026Mar 27, 2026Apr 2, 2026Apr 8, 2026Apr 15, 2026

Evaluation results

MethodLinks
MDPDBackbone=BERT+BUTD, Params. (M)=5.9, Mem. (G)=3.4, QPS=24.052026.0481.961.1—516.2
UniPTBackbone=BERT+BUTD, Params. (M)=5.9, Mem. (G)=3.1, QPS=12.752026.0480.259.8—510.5
SSFBackbone=BERT+BUTD, Params. (M)=0.2, Mem. (G)=8.4, QPS=22.232026.048060.4—512.8
AdaLoRABackbone=BERT+BUTD, Params. (M)=1.0, Mem. (G)=8.8, QPS=23.032026.0479.860.1—510.3
Fully-FTBackbone=BERT+BUTD, Params. (M)=109.5, Mem. (G)=9.9, QPS=24.142026.0479.762.1—513.5
FacTBackbone=BERT+BUTD, Params. (M)=0.6, Mem. (G)=8.72026.0479.259.3—508.8
AdapterBackbone=BERT+BUTD, Params. (M)=2.6, Mem. (G)=8.8, QPS=22.442026.0479.160.5—511.3
LoRABackbone=BERT+BUTD, Params. (M)=1.1, Mem. (G)=8.8, QPS=23.372026.0478.859.6—508.2
PromptBackbone=BERT+BUTD, Params. (M)=10.7, Mem. (G)=9.4, QPS=17.642026.0478.759—508.5
LSTBackbone=BERT+BUTD, Params. (M)=7.5, Mem. (G)=4.6, QPS=10.732026.0477.957.3—501.9
BitFitBackbone=BERT+BUTD, Params. (M)=0.9, Mem. (G)=8.6, QPS=23.702026.0477.357.8—503.9
PartiallyBackbone=BERT+BUTD, Params. (M)=0.8, Mem. (G)=1.0, QPS=24.192026.0474.857.3—485.5
Cluster-aware UpcyclingArch.=MoE, Backbone=ViT-B/16, Samples=4.0B+1.3B, Zero-shot protocol=true2026.0435.451.643.5—
Sparse UpcyclingArch.=MoE, Backbone=ViT-B/16, Samples=4.0B+1.3B, Zero-shot protocol=true2026.0434.950.942.9—
DenseArch.=Dense, Backbone=ViT-B/16, Samples=4.0B+1.3B, Zero-shot protocol=true2026.0434.350.842.6—
Drop-UpcyclingArch.=MoE, Backbone=ViT-B/16, Samples=4.0B+1.3B, Zero-shot protocol=true2026.0434.151.342.7—
CLIP-MoEArch.=MoE, Backbone=ViT-B/16, Samples=4.0B+1.3B, Zero-shot protocol=true2026.043451.542.8—
DenseArch.=Dense, Backbone=ViT-B/16, Samples=4.0B, Zero-shot protocol=true2026.0432.549.140.8—
DeRS-LMArch.=MoE, Backbone=ViT-B/32, Samples=4.0B+1.3B, Zero-shot protocol=true2026.043147.739.4—
Cluster-aware UpcyclingArch.=MoE, Backbone=ViT-B/32, Samples=4.0B+1.3B, Zero-shot protocol=true2026.043148.239.6—
DenseArch.=Dense, Backbone=ViT-B/32, Samples=4.0B+1.3B, Zero-shot protocol=true2026.0430.847.539.2—
Sparse UpcyclingArch.=MoE, Backbone=ViT-B/32, Samples=4.0B+1.3B, Zero-shot protocol=true2026.0430.84839.4—
Drop-UpcyclingArch.=MoE, Backbone=ViT-B/32, Samples=4.0B+1.3B, Zero-shot protocol=true2026.0429.746.538.1—
CLIP-MoEArch.=MoE, Backbone=ViT-B/32, Samples=4.0B+1.3B, Zero-shot protocol=true2026.0429.546.838.2—
DenseArch.=Dense, Backbone=ViT-B/32, Samples=4.0B, Zero-shot protocol=true2026.0425.542.534—
CLIPBackbone=RN50, Evaluation Protocol=zero-shot, Pre-training Dataset=CC3M2026.0321.326.924.1—
CLIP + MM-TSBackbone=RN50, Evaluation Protocol=zero-shot, Pre-training Dataset=CC3M2026.0321.228.424.8—