Fine-Grained Image Classification on Stanford Cars (Base/New/H Metrics)
90Base AccuracyToolFG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ToolFGTraining Strategy=per-dataset training, Evaluation Protocol=Zero-shot base-to-novel generalization2026.06 | 90 | 90.5 | 90.2 | — | |
| GAPL2026.04 | 84.7 | 74.5 | 79.27 | — | |
| HiCroPL2026.04 | 81.51 | 75.04 | 78.14 | — | |
| MMRL2026.04 | 81.3 | 75.07 | 78.06 | — | |
| TCP2026.04 | 80.8 | 74.13 | 77.32 | — | |
| MMA2026.04 | 78.5 | 73.1 | 75.7 | — | |
| ProMetaRShots=162024.04 | 78.32 | 75.18 | 76.72 | — | |
| PromptSRCTraining Strategy=per-dataset training, Evaluation Protocol=Zero-shot base-to-novel generalization2026.06 | 78.3 | 75 | 76.6 | — | |
| PromptSRCShots=162024.04 | 78.27 | 74.97 | 76.58 | — | |
| PromptSRC2026.04 | 78.27 | 74.97 | 76.58 | — | |
| CoOpShots=162024.04 | 78.12 | 60.4 | 68.13 | — | |
| CoOp2026.04 | 78.12 | 60.4 | 68.13 | — | |
| CoPrompt2026.04 | 76.97 | 74.4 | 75.66 | — | |
| RPOShots=162024.04 | 73.87 | 75.53 | 74.69 | — | |
| UNIGRAMShots=162024.04 | 73.5 | 75.38 | 74.43 | — | |
| MaPLeShots=162024.04 | 72.94 | 74 | 73.47 | — | |
| MaPLe2026.04 | 72.94 | 74 | 73.47 | — | |
| MaPLeTraining Strategy=per-dataset training, Evaluation Protocol=Zero-shot base-to-novel generalization2026.06 | 72.9 | 74 | 73.5 | — | |
| IVLPShots=162024.04 | 72.39 | 73.31 | 72.85 | — | |
| KgCoOp2026.04 | 71.76 | 75.04 | 73.36 | — | |
| CoCoOpShots=162024.04 | 70.49 | 73.59 | 72.01 | — | |
| CoCoOp2026.04 | 70.49 | 73.59 | 72.01 | — | |
| DiVE-kTraining Strategy=per-dataset training, Evaluation Protocol=Zero-shot base-to-novel generalization2026.06 | 69 | 76.2 | 72.4 | — | |
| ViRFTTraining Strategy=per-dataset training, Evaluation Protocol=Zero-shot base-to-novel generalization2026.06 | 64.8 | 76.9 | 70.4 | — | |
| Fine-R1Training Strategy=per-dataset training, Evaluation Protocol=Zero-shot base-to-novel generalization2026.06 | 63.6 | 75.1 | 68.9 | — | |
| CLIPShots=Zero-shot2024.04 | 63.37 | 74.89 | 68.65 | — | |
| Qwen2.5-VL-7BTraining Strategy=per-dataset training, Evaluation Protocol=Zero-shot base-to-novel generalization2026.06 | 58.9 | 72.9 | 65.1 | — | |
| CLIPProtocol=Base-to-novel generalization2025.12 | — | — | — | 63.37 | |
| CoCoOpProtocol=Base-to-novel generalization2025.12 | — | — | — | 70.49 | |
| CoOpProtocol=Base-to-novel generalization2025.12 | — | — | — | 75.67 | |
| FARLProtocol=Base-to-novel generalization2025.12 | — | — | — | 81.43 | |
| KgCoOpProtocol=Base-to-novel generalization2025.12 | — | — | — | 71.76 | |
| MaPLeProtocol=Base-to-novel generalization2025.12 | — | — | — | 72.94 | |
| MetaPromptProtocol=Base-to-novel generalization2025.12 | — | — | — | 76.34 | |
| MMAProtocol=Base-to-novel generalization2025.12 | — | — | — | 78.5 | |
| MMRLProtocol=Base-to-novel generalization2025.12 | — | — | — | 81.5 | |
| PLOTProtocol=Base-to-novel generalization2025.12 | — | — | — | 79.07 | |
| ProVPProtocol=Base-to-novel generalization2025.12 | — | — | — | 80.43 | |
| TCPProtocol=Base-to-novel generalization2025.12 | — | — | — | 80.8 |