Image Classification on iNaturalist (accuracy %)
73.8AccuracyiSICE
Evaluation results
| Method | Links | |
|---|---|---|
| iSICEBackbone=ResNeXt-1012023.04 | 73.8 | |
| iSICEBackbone=Swin-B2023.04 | 72.4 | |
| Precision ΩBackbone=ResNeXt-1012023.04 | 72.3 | |
| iSQRT-COVBackbone=ResNeXt-1012023.04 | 72.2 | |
| iSICEBackbone=ResNet-1012023.04 | 72 | |
| iSQRT-COVBackbone=ResNet-1012023.04 | 70.6 | |
| Precision ΩBackbone=ResNet-1012023.04 | 69.8 | |
| iSQRT-COVBackbone=Swin-B2023.04 | 69.7 | |
| Direct FTMethod Category=PT, Backbone=DeiT-S, 11.3M, Para. (M)=11.02024.06 | 69.3 | |
| FRONTBackbone=ResNet-1522026.03 | 69.1 | |
| FRONTBackbone=ResNet-502026.03 | 69 | |
| WAVEMethod Category=Learngene, Backbone=DeiT-S, 11.3M, Para. (M)=4.02024.06 | 67.6 | |
| Precision ΩBackbone=Swin-B2023.04 | 67.3 | |
| LiGOBackbone=DeiT-B2026.03 | 67.1 | |
| MangoBackbone=DeiT-B2026.03 | 67.1 | |
| BoTBackbone=DeiT-B2026.03 | 67.1 | |
| Auto-LGMethod Category=Learngene, Backbone=DeiT-S, 11.3M, Para. (M)=7.52024.06 | 67 | |
| Direct FTMethod Category=PT, Backbone=DeiT-Ti, 3.0M, Para. (M)=2.92024.06 | 66.9 | |
| TLEGMethod Category=Learngene, Backbone=DeiT-S, 11.3M, Para. (M)=3.92024.06 | 66.5 | |
| LIGOMethod Category=Trans., Backbone=DeiT-S, 11.3M, Para. (M)=7.52024.06 | 66.1 | |
| WAVEMethod Category=Learngene, Backbone=DeiT-Ti, 3.0M, Para. (M)=1.12024.06 | 65.2 | |
| iSICEBackbone=ConvNext-T2023.04 | 65 | |
| TLEGMethod Category=Learngene, Backbone=DeiT-Ti, 3.0M, Para. (M)=1.12024.06 | 63.4 | |
| Share InitMethod Category=Trans., Backbone=DeiT-Ti, 3.0M, Para. (M)=0.62024.06 | 63.2 | |
| Share InitMethod Category=Trans., Backbone=DeiT-S, 11.3M, Para. (M)=2.22024.06 | 63 | |
| Random InitializationBackbone=ResNet-1522026.03 | 62.7 | |
| Auto-LGMethod Category=Learngene, Backbone=DeiT-Ti, 3.0M, Para. (M)=2.02024.06 | 62.5 | |
| iSICEBackbone=Swin-T2023.04 | 61.9 | |
| BoTBackbone=DeiT-S2026.03 | 61.9 | |
| Random InitializationBackbone=ResNet-502026.03 | 61.8 | |
| LIGOMethod Category=Trans., Backbone=DeiT-Ti, 3.0M, Para. (M)=2.02024.06 | 61.6 | |
| iSQRT-COVBackbone=ConvNext-T2023.04 | 61.5 | |
| WSBackbone=DeiT-S2026.03 | 61.2 | |
| bert2BERTBackbone=DeiT-B2026.03 | 61.1 | |
| iSICEBackbone=ResNet-502023.04 | 60.7 | |
| +GRPOCategory=DeepTaxon, Model=Qwen2.5-VL-7B2026.04 | 60.07 | |
| iSICEBackbone=VGG-162023.04 | 59.8 | |
| Precision ΩBackbone=ResNet-502023.04 | 59.6 | |
| iSICEBackbone=VGG-192023.04 | 59.6 | |
| Precision ΩBackbone=ConvNext-T2023.04 | 59.3 | |
| Heur-LGMethod Category=Learngene, Backbone=DeiT-S, 11.3M, Para. (M)=5.72024.06 | 59.3 | |
| Precision ΩBackbone=Swin-T2023.04 | 59.1 | |
| ScratchBackbone=DeiT-B2026.03 | 58.2 | |
| iSQRT-COVBackbone=Swin-T2023.04 | 58.1 | |
| Precision ΩBackbone=VGG-162023.04 | 57.9 | |
| +SFTCategory=DeepTaxon, Model=Qwen2.5-VL-7B2026.04 | 57.8 | |
| iSQRT-COVBackbone=ResNet-502023.04 | 57.7 | |
| Heur-LGMethod Category=Learngene, Backbone=DeiT-Ti, 3.0M, Para. (M)=1.52024.06 | 57.4 | |
| Precision ΩBackbone=VGG-192023.04 | 57.3 | |
| iSQRT-COVBackbone=VGG-192023.04 | 56.9 | |
| iSQRT-COVBackbone=VGG-162023.04 | 56.2 | |
| GHN-3Method Category=Direct, Backbone=DeiT-S, 11.3M, Para. (M)=02024.06 | 55.5 | |
| Wt SelectMethod Category=Trans., Backbone=DeiT-S, 11.3M, Para. (M)=11.02024.06 | 54.2 | |
| He-InitMethod Category=Direct, Backbone=DeiT-S, 11.3M, Para. (M)=02024.06 | 54 | |
| ScratchBackbone=DeiT-S2026.03 | 54 | |
| He-InitMethod Category=Direct, Backbone=DeiT-Ti, 3.0M, Para. (M)=02024.06 | 52.3 | |
| MimeticMethod Category=Direct, Backbone=DeiT-S, 11.3M, Para. (M)=02024.06 | 52.2 | |
| GHN-3Method Category=Direct, Backbone=DeiT-Ti, 3.0M, Para. (M)=02024.06 | 51.7 | |
| Wt SelectMethod Category=Trans., Backbone=DeiT-Ti, 3.0M, Para. (M)=2.92024.06 | 51.2 | |
| MimeticMethod Category=Direct, Backbone=DeiT-Ti, 3.0M, Para. (M)=02024.06 | 49 | |
| MSP (Cls-Preserving)Category=OOD (threshold), Model=ViT-H2026.04 | 39.3 | |
| VIM (Cls-Preserving)Category=OOD (threshold), Model=ViT-H2026.04 | 39.3 | |
| MSP (Disc-Optimized†)Category=OOD (threshold), Model=ViT-H2026.04 | 38.24 | |
| VIM (Disc-Optimized†)Category=OOD (threshold), Model=ViT-H2026.04 | 38.09 | |
| Zero-Shot‡Category=DeepTaxon, Model=Qwen2.5-VL-7B2026.04 | 24.87 | |
| OracleScenario=Scenario A, Backbone=ConvNext-Large2025.12 | 21.4 | |
| Neural Coherence (NC@5)Scenario=Scenario A, Backbone=ConvNext-Large2025.12 | 21.13 | |
| MSP (Cls-Preserving)Category=OOD (threshold), Model=ResNet-502026.04 | 19.92 | |
| VIM (Cls-Preserving)Category=OOD (threshold), Model=ResNet-502026.04 | 19.92 | |
| Source-ValScenario=Scenario A, Backbone=ConvNext-Large2025.12 | 19.46 | |
| MSP (Disc-Optimized†)Category=OOD (threshold), Model=ResNet-502026.04 | 19.07 | |
| VIM (Disc-Optimized†)Category=OOD (threshold), Model=ResNet-502026.04 | 18.93 | |
| ABEScenario=Scenario A, Backbone=ConvNext-Large2025.12 | 17.23 | |
| PARCScenario=Scenario A, Backbone=ConvNext-Large2025.12 | 0.07 |