Image Classification on Places365
55.66AccuracyStandard
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| StandardBackbone=ViT-L/142025.11 | 55.66 | — | — | — | — | |
| StandardBackbone=ViT-L/142025.11 | 55.66 | — | — | — | — | |
| CBM-SuiteBackbone=Perception Encoder, Selection=Best-performing vision backbone–VLM pair2026.03 | 54.64 | — | — | — | — | |
| StandardBackbone=ResNet502025.11 | 53.28 | — | — | — | — | |
| StandardBackbone=ResNet502025.11 | 53.28 | — | — | — | — | |
| FCBMBackbone=ViT-L/14, Average NEC=28.082025.11 | 51.39 | — | — | — | — | |
| FCBMBackbone=ViT-L/142025.11 | 51.39 | — | — | — | — | |
| LF-CBMBackbone=ViT-L/14, Average NEC=27.852025.11 | 50.51 | — | — | — | — | |
| LF-CBMBackbone=ViT-L/142025.11 | 50.51 | — | — | — | — | |
| FCBMBackbone=ResNet50, Average NEC=28.852025.11 | 49.13 | — | — | — | — | |
| FCBMBackbone=ResNet502025.11 | 49.13 | — | — | — | — | |
| LF-CBMBackbone=ResNet50, Average NEC=27.372025.11 | 48.88 | — | — | — | — | |
| LF-CBMBackbone=ResNet502025.11 | 48.88 | — | — | — | — | |
| CF-CBMBackbone=ViT-L/14, Average NEC=28.562025.11 | 48.55 | — | — | — | — | |
| CF-CBMBackbone=ViT-L/142025.11 | 48.55 | — | — | — | — | |
| LAGOBackbone=L/14, Zero-shot=true2026.05 | 47.28 | — | — | — | — | |
| CF-CBMBackbone=ResNet50, Average NEC=28.462025.11 | 46.39 | — | — | — | — | |
| CF-CBMBackbone=ResNet502025.11 | 46.39 | — | — | — | — | |
| LAGOBackbone=B/16, Zero-shot=true2026.05 | 45.56 | — | — | — | — | |
| Standard (sparse)Backbone=ViT-L/14, Average NEC=28.922025.11 | 44.9 | — | — | — | — | |
| Standard (sparse)Backbone=ViT-L/142025.11 | 44.9 | — | — | — | — | |
| Standard (sparse)Backbone=ResNet50, Average NEC=29.842025.11 | 44 | — | — | — | — | |
| Standard (sparse)Backbone=ResNet502025.11 | 44 | — | — | — | — | |
| LFCBMSelection=Best-performing vision backbone–VLM pair2026.03 | 43.68 | — | — | — | — | |
| PCBMBackbone=ViT-L/14, Average NEC=32.562025.11 | 43.18 | — | — | — | — | |
| PCBMBackbone=ViT-L/142025.11 | 43.18 | — | — | — | — | |
| BiFTABackbone=L/14, Zero-shot=true2026.05 | 42.98 | — | — | — | — | |
| LAGOBackbone=B/32, Zero-shot=true2026.05 | 42.59 | — | — | — | — | |
| PCBMBackbone=ResNet50, Average NEC=32.372025.11 | 42.26 | — | — | — | — | |
| PCBMBackbone=ResNet502025.11 | 42.26 | — | — | — | — | |
| WCABackbone=L/14, Zero-shot=true2026.05 | 42.23 | — | — | — | — | |
| BiFTABackbone=B/16, Zero-shot=true2026.05 | 42.12 | — | — | — | — | |
| BiFTABackbone=B/32, Zero-shot=true2026.05 | 41.55 | — | — | — | — | |
| WCABackbone=B/16, Zero-shot=true2026.05 | 41.43 | — | — | — | — | |
| WaffleBackbone=L/14, Zero-shot=true2026.05 | 40.89 | — | — | — | — | |
| CuPLBackbone=L/14, Zero-shot=true2026.05 | 40.77 | — | — | — | — | |
| WaffleBackbone=B/16, Zero-shot=true2026.05 | 40.76 | — | — | — | — | |
| WCABackbone=B/32, Zero-shot=true2026.05 | 40.66 | — | — | — | — | |
| CLIP-EBackbone=L/14, Zero-shot=true2026.05 | 40.55 | — | — | — | — | |
| CLIP-DBackbone=L/14, Zero-shot=true2026.05 | 40.55 | — | — | — | — | |
| Whole TMode=Whole Dataset2026.03 | 40.4 | — | — | — | — | |
| CLIP-DBackbone=B/16, Zero-shot=true2026.05 | 40.34 | — | — | — | — | |
| CLIP-EBackbone=B/16, Zero-shot=true2026.05 | 40.3 | — | — | — | — | |
| CLIP-DBackbone=B/32, Zero-shot=true2026.05 | 39.9 | — | — | — | — | |
| CuPLBackbone=B/16, Zero-shot=true2026.05 | 39.83 | — | — | — | — | |
| CLIPBackbone=L/14, Zero-shot=true2026.05 | 39.63 | — | — | — | — | |
| WaffleBackbone=B/32, Zero-shot=true2026.05 | 39.47 | — | — | — | — | |
| CLIP-EBackbone=B/32, Zero-shot=true2026.05 | 39.28 | — | — | — | — | |
| CLIPBackbone=B/16, Zero-shot=true2026.05 | 39.27 | — | — | — | — | |
| CuPLBackbone=B/32, Zero-shot=true2026.05 | 39.08 | — | — | — | — | |
| CLIP ViT-B/16Alignment train data=WIT400M, Zero-shot=true2026.05 | 38.96 | — | — | — | — | |
| CLIPBackbone=B/32, Zero-shot=true2026.05 | 38.51 | — | — | — | — | |
| CoDyRAZero-shot=true2024.12 | 36.52 | — | — | — | — | |
| H8(ViT-H/16)FLOPs=77.5 x 10^9, Evaluation Protocol=linear2025.05 | 35.1 | — | — | — | — | |
| BEiT-B/16Alignment train data=IN21K repr. + 1 img-capt. pair per class, Zero-shot=true2026.05 | 34.78 | — | — | — | — | |
| ViT-H/16FLOPs=127.7 x 10^9, Evaluation Protocol=linear2025.05 | 34.7 | — | — | — | — | |
| BEiT-B/16Alignment train data=IN21K repr., Zero-shot=true2026.05 | 34.34 | — | — | — | — | |
| I8(ViT-H/16)FLOPs=77.7 x 10^9, Evaluation Protocol=linear2025.05 | 33.9 | — | — | — | — | |
| Zero-ShotZero-shot=true2024.12 | 33.77 | — | — | — | — | |
| MoE-AdapterZero-shot=true2024.12 | 33.77 | — | — | — | — | |
| RAILZero-shot=true2024.12 | 33.77 | — | — | — | — | |
| H8(ViT-L/16)FLOPs=37.7 x 10^9, Evaluation Protocol=linear2025.05 | 33.6 | — | — | — | — | |
| ViT-L/16FLOPs=61.9 x 10^9, Evaluation Protocol=linear2025.05 | 33.2 | — | — | — | — | |
| I8(ViT-L/16)FLOPs=38.5 x 10^9, Evaluation Protocol=linear2025.05 | 32.6 | — | — | — | — | |
| H8(ViT-H/16)FLOPs=77.5 x 10^9, Evaluation Protocol=k-NN2025.05 | 31.1 | — | — | — | — | |
| ViT-H/16FLOPs=127.7 x 10^9, Evaluation Protocol=k-NN2025.05 | 30.6 | — | — | — | — | |
| H8(ViT-L/16)FLOPs=37.7 x 10^9, Evaluation Protocol=k-NN2025.05 | 29.4 | — | — | — | — | |
| I8(ViT-H/16)FLOPs=77.7 x 10^9, Evaluation Protocol=k-NN2025.05 | 29.2 | — | — | — | — | |
| ViT-L/16FLOPs=61.9 x 10^9, Evaluation Protocol=k-NN2025.05 | 28.4 | — | — | — | — | |
| I8(ViT-L/16)FLOPs=38.5 x 10^9, Evaluation Protocol=k-NN2025.05 | 28 | — | — | — | — | |
| BEiT-B/16Alignment train data=IN1K repr., Zero-shot=true2026.05 | 24.32 | — | — | — | — | |
| DM+OursIPC=102026.03 | 19.2 | — | — | — | — | |
| IDCIPC=102026.03 | 9.8 | — | — | — | — | |
| DM+OursIPC=12026.03 | 7.7 | — | — | — | — | |
| DMIPC=102026.03 | 5.8 | — | — | — | — | |
| IDCIPC=12026.03 | 4 | — | — | — | — | |
| RandomIPC=102026.03 | 3.9 | — | — | — | — | |
| DMIPC=12026.03 | 2.6 | — | — | — | — | |
| RandomIPC=12026.03 | 0.9 | — | — | — | — | |
| CEMbackbone=ViT, concept_embeddings=Label-Free CBM, training_strategy=independent2026.06 | — | — | — | 45.01 | — | |
| ClassifierBackbone=DeiT-Ti2022.03 | — | — | — | 40.36 | — | |
| ClassifierBackbone=DeiT-S2022.03 | — | — | — | 44.87 | — | |
| ClassifierBackbone=T2T-ViT-122022.03 | — | — | — | 37.06 | — | |
| Concept Matrix Search2024.04 | — | — | — | 39.43 | — | |
| Contrastive-CBM2024.04 | — | — | — | 40.22 | — | |
| CoopCBMbackbone=ViT, concept_embeddings=Label-Free CBM, training_strategy=independent2026.06 | — | — | — | 48.21 | — | |
| DeiT-TFLOPS(G)=1.07, Param(M)=5.562023.04 | — | — | — | 36.33 | 66.72 | |
| DescriptionCLS2024.04 | — | — | — | 40.55 | — | |
| DOT-CBMbackbone=ViT, concept_embeddings=Label-Free CBM, training_strategy=independent2026.06 | — | — | — | 50.65 | — | |
| Hiera-Bbackbone=Hiera-B2023.06 | — | — | — | 58.9 | — | |
| Hiera-B+backbone=Hiera-B+2023.06 | — | — | — | 59.2 | — | |
| Hiera-Hbackbone=Hiera-H2023.06 | — | — | — | 60 | — | |
| Hiera-H448backbone=Hiera-H4482023.06 | — | — | — | 60.6 | — | |
| Hiera-Lbackbone=Hiera-L2023.06 | — | — | — | 59.6 | — | |
| ImageBindSetting=Fine-tuning, Protocol=linear2024.03 | — | — | — | 49.45 | — | |
| ImageBindSetting=Zero-shot2024.03 | — | — | — | 45.4 | — | |
| InceptionNeXt-AFLOPS(G)=0.5, Param(M)=3.542023.04 | — | — | — | 35.67 | 65.78 | |
| IndividualBackbone=DeiT-Ti2022.03 | — | — | — | 52.67 | — | |
| IndividualBackbone=DeiT-S2022.03 | — | — | — | 55.22 | — | |
| IndividualBackbone=T2T-ViT-122022.03 | — | — | — | 51.53 | — |