Image Classification on SUN397 (Accuracy, ECE)
82.4AccuracyθB fine-tune
Evaluation results
| Method | Links | ||
|---|---|---|---|
| θB fine-tuneTarget Backbone=ViT-L/14, Target Pre-training=Datacomp-XL, Evaluation Protocol=fine-tune2026.05 | 82.4 | — | |
| θA fine-tuneNumber of calibration samples (N)=-2026.05 | 82.4 | — | |
| θA fine-tuneSource Backbone=ViT-B/16, Source Pre-training=LAION-2B, Evaluation Protocol=fine-tune2026.05 | 79.35 | — | |
| theta_A fine-tuneN=–2026.05 | 79.25 | — | |
| θB fine-tuneNumber of calibration samples (N)=-2026.05 | 79.25 | — | |
| GalLoP+OursShots=42026.05 | 78.4 | — | |
| theta_B fine-tuneN=–2026.05 | 78.29 | — | |
| PromptSRC+OursShots=42026.05 | 77.6 | — | |
| SCANShot=162025.05 | 77.32 | — | |
| EBMmode=Fine-tuning2025.10 | 77.3 | — | |
| PromptSRCShots=42026.05 | 77.2 | — | |
| TCP+OursShots=42026.05 | 77.2 | — | |
| GalLoPShots=42026.05 | 77.2 | — | |
| PromptSRCmode=Fine-tuning2025.10 | 77.2 | — | |
| TCP*Shots=42026.05 | 76.4 | — | |
| PLOTShots=42026.05 | 76 | — | |
| PLOTmode=Fine-tuning2025.10 | 76 | — | |
| ProDAShots=42026.05 | 75.7 | — | |
| MaPLeShots=42026.05 | 75.5 | — | |
| MaPLemode=Fine-tuning2025.10 | 75.5 | — | |
| SCANShot=82025.05 | 75.43 | — | |
| BiCoSource Backbone=ViT-B/16, Source Pre-training=LAION-2B, Target Backbone=ViT-L/14, Target Pre-training=Datacomp-XL, N=100, Evaluation Protocol=task-vector transfer2026.05 | 74.74 | — | |
| CoOpShots=42026.05 | 74.7 | — | |
| CoOpmode=Fine-tuning2025.10 | 74.7 | — | |
| LoCoOpShots=42026.05 | 74.2 | — | |
| THESEUSSource Backbone=ViT-B/16, Source Pre-training=LAION-2B, Target Backbone=ViT-L/14, Target Pre-training=Datacomp-XL, N=100, Evaluation Protocol=task-vector transfer2026.05 | 73.97 | — | |
| θB + τA paddedSource Backbone=ViT-B/16, Source Pre-training=LAION-2B, Target Backbone=ViT-L/14, Target Pre-training=Datacomp-XL, Evaluation Protocol=task-vector transfer2026.05 | 73.93 | — | |
| θB zero-shotTarget Backbone=ViT-L/14, Target Pre-training=Datacomp-XL, Evaluation Protocol=zero-shot2026.05 | 73.77 | — | |
| CLIP-LoRA+OursShots=42026.05 | 73.5 | — | |
| CLIP-LoRAShots=42026.05 | 72.5 | — | |
| SCANShot=42025.05 | 72.49 | — | |
| SimNLShot=162025.05 | 72.32 | — | |
| CoCoOpShots=42026.05 | 72.2 | — | |
| CoCoOpmode=Fine-tuning2025.10 | 72.2 | — | |
| Tip-Adapter-FShot=162025.05 | 71.31 | — | |
| SimNLShot=82025.05 | 70.93 | — | |
| TaskResShot=162025.05 | 70.67 | — | |
| BiCoNumber of calibration samples (N)=1002026.05 | 70.12 | — | |
| θB zero-shotNumber of calibration samples (N)=-2026.05 | 70.03 | — | |
| THESEUSNumber of calibration samples (N)=1002026.05 | 69.97 | — | |
| SCANShot=22025.05 | 69.83 | — | |
| θB + τA cropNumber of calibration samples (N)=-2026.05 | 69.71 | — | |
| CLIP-AdapterShot=162025.05 | 69.59 | — | |
| BiCoN=1002026.05 | 69.46 | — | |
| CoOpShot=162025.05 | 69.11 | — | |
| BiCoN=502026.05 | 69.09 | — | |
| TIP-Adapter-F+OursShots=42026.05 | 68.9 | — | |
| THESEUSN=1002026.05 | 68.83 | — | |
| BiCoN=202026.05 | 68.81 | — | |
| Tip-Adapter-FShot=82025.05 | 68.8 | — | |
| TaskResShot=82025.05 | 68.73 | — | |
| THESEUSN=502026.05 | 68.73 | — | |
| SimNLShot=42025.05 | 68.59 | — | |
| BiCoN=102026.05 | 68.58 | — | |
| THESEUSN=102026.05 | 68.55 | — | |
| theta_B zero-shotN=–2026.05 | 68.5 | — | |
| THESEUSN=202026.05 | 68.44 | — | |
| MaPLe+FPPBackbone=CLIP-ViT/B16, Prompt Initialization Source=MaPLe, Adaptation Framework=TPT, Adaptation Strategy=FPP2026.04 | 68.29 | 1.74 | |
| MMA + NeRP2026.05 | 68.25 | — | |
| MMA2026.05 | 68.17 | — | |
| RAR (LLaVA1.5)Shots=162024.03 | 68 | — | |
| theta_B + tau_A^paddedN=–2026.05 | 68 | — | |
| SCANShot=12025.05 | 67.93 | — | |
| MMRL + NeRP2026.05 | 67.73 | — | |
| MMRL++Backbone=ViT-B/16, Training Shots=16-shot2026.05 | 67.65 | — | |
| MMRL2026.05 | 67.57 | — | |
| CLIP-AdapterShot=82025.05 | 67.52 | — | |
| TIP-Adapter-FShots=42026.05 | 67.5 | — | |
| CoCoOp + NeRP2026.05 | 67.43 | — | |
| CoCoOp2026.05 | 67.36 | — | |
| PromptSRC + NeRP2026.05 | 67.3 | — | |
| TaskResShot=42025.05 | 67.27 | — | |
| MMRLBackbone=ViT-B/16, Training Shots=16-shot2026.05 | 67.24 | — | |
| CoCoOpShot=162025.05 | 67.21 | — | |
| A3B2Backbone=ViT-B/16, Training Shots=16-shot2026.05 | 67.15 | — | |
| TCPBackbone=ViT-B/16, Training Shots=16-shot2026.05 | 67.13 | — | |
| PromptSRC2026.05 | 67.1 | — | |
| KgCoOpBackbone=ViT-B/16, Training Shots=16-shot2026.05 | 67.03 | — | |
| SimNLShot=22025.05 | 66.73 | — | |
| CoOp+FPPBackbone=CLIP-ViT/B16, Prompt Initialization Source=CoOp, Adaptation Framework=TPT, Adaptation Strategy=FPP2026.04 | 66.44 | 3 | |
| CoOpBackbone=ViT-B/16, Training Shots=16-shot2026.05 | 66.36 | — | |
| MMABackbone=ViT-B/16, Training Shots=16-shot2026.05 | 66.32 | — | |
| Tip-Adapter-FShot=42025.05 | 66.13 | — | |
| MaPLeBackbone=ViT-B/16, Training Shots=16-shot2026.05 | 66.04 | — | |
| CLIP-AdapterShot=42025.05 | 65.9 | — | |
| CoOp+TPTBackbone=CLIP-ViT/B16, Prompt Initialization Source=CoOp, Adaptation Framework=TPT, Adaptation Strategy=TPT2026.04 | 65.6 | 20.8 | |
| CoOpShot=82025.05 | 65.43 | — | |
| CoCoOpShot=82025.05 | 65.37 | — | |
| SimNLShot=12025.05 | 65.13 | — | |
| CoOp+C-TPTBackbone=CLIP-ViT/B16, Prompt Initialization Source=CoOp, Adaptation Framework=TPT, Adaptation Strategy=C-TPT2026.04 | 65.1 | 11.8 | |
| TaskResShot=22025.05 | 64.97 | — | |
| CoCoOpShot=42025.05 | 64.9 | — | |
| MaPLe+TPTBackbone=CLIP-ViT/B16, Prompt Initialization Source=MaPLe, Adaptation Framework=TPT, Adaptation Strategy=TPT2026.04 | 64.87 | 8.47 | |
| CoOp+O-TPTBackbone=CLIP-ViT/B16, Prompt Initialization Source=CoOp, Adaptation Framework=TPT, Adaptation Strategy=O-TPT2026.04 | 64.01 | 7.23 | |
| Tip-Adapter-FShot=22025.05 | 63.55 | — | |
| CoCoOpShot=22025.05 | 63.5 | — | |
| MaPLe+O-TPTBackbone=CLIP-ViT/B16, Prompt Initialization Source=MaPLe, Adaptation Framework=TPT, Adaptation Strategy=O-TPT2026.04 | 63.49 | 2.78 | |
| CoOpShot=42025.05 | 63.46 | — | |
| RAR (LLaVA1.5)Shots=82024.03 | 63.4 | — | |
| Tip-Adapter-FShot=12025.05 | 62.51 | — |