Click-through rate prediction on Industrial 7 billion user interaction records (test)
0.6994AUCMGDIN
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MGDINlayers=3, windows=4, granularities={32, 64, 96, 128}, layer-wise sparsity ratios={0.33, 0.66, 1.0}2026.03 | 0.6994 | 0.5936 | 0.1116 | |
| RankMixer2026.03 | 0.6956 | 0.5916 | 0.1118 | |
| OneTrans2026.03 | 0.6956 | 0.5921 | 0.1118 | |
| DCN2026.03 | 0.6915 | 0.5914 | 0.1121 | |
| GDCN2026.03 | 0.6894 | 0.5899 | 0.1123 | |
| DNN2026.03 | 0.6888 | 0.5901 | 0.1122 | |
| MaskNet2026.03 | 0.6881 | 0.5876 | 0.1123 | |
| Wide&Deep2026.03 | 0.6875 | 0.5888 | 0.1124 | |
| AutoInt2026.03 | 0.6874 | 0.5877 | 0.1124 | |
| FM2026.03 | 0.6857 | 0.5893 | 0.1125 | |
| DeepFM2026.03 | 0.6848 | 0.586 | 0.1125 |