Concept-based Classification 벤치마크
Concept-based Classification on AwA2
Task Accuracy (%)
- 2024-08-13 — CGEM (ResNet-34): Task Accuracy (%) 94.63
- 2024-09-22 — EQ-CBM (ResNet-34): Task Accuracy (%) 95.965
| Rank | Model | Task Accuracy (%) | Concept Accuracy (%) | Paper | Code | Year |
|---|---|---|---|---|---|---|
| 1 | EQ-CBM (ResNet-34) | 95.965 | 99.129 | EQ-CBM: A Probabilistic Concept Bottleneck with Energy-based Models and Quantized Vectors | 2024 | |
| 2 | CGEM (ResNet-34) | 94.63 | 93.68 | Concept Graph Embedding Models for Enhanced Accuracy and Interpretability | jumpsnack/cgem | 2024 |
| 3 | EQ-CBM (ResNet-34) | 95.965 | 99.129 | EQ-CBM: A Probabilistic Concept Bottleneck with Energy-based Models and Quantized Vectors | 2024 | |
| 4 | CGEM (ResNet-34) | 94.63 | 93.68 | Concept Graph Embedding Models for Enhanced Accuracy and Interpretability | jumpsnack/cgem | 2024 |