Medical Image Classification on Medical Fairness Benchmark Aggregate (test)
11.39PPGRPO
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GRPOBackbone Model=Qwen-2.5-VL-7B2025.10 | 11.39 | 9.09 | 4.61 | 0.0463 | 0.0973 | 4.68 | 9.43 | 80.45 | 25.5 | 76.85 | 24.37 | |
| GRPO+DROBackbone Model=Qwen-2.5-VL-7B2025.10 | 14.51 | 7.41 | 7.42 | 0.0326 | 0.0654 | 5.62 | 11.5 | 75.1 | 25.86 | 71.1 | 25.04 | |
| Re++Backbone Model=Qwen-2.5-VL-7B2025.10 | 15.18 | 7.79 | 6.23 | 0.0322 | 0.065 | 4.71 | 9.61 | 75.32 | 26.12 | 71.93 | 25.31 | |
| EQPOBackbone Model=Qwen-2.5-VL-7B2025.10 | 16.8 | 5.55 | 4.39 | 0.0229 | 0.0452 | 4.41 | 8.93 | 80.75 | 26.47 | 77.34 | 25.88 | |
| GRPO+RSBackbone Model=Qwen-2.5-VL-7B2025.10 | 21.56 | 8.09 | 4.96 | 0.0316 | 0.0636 | 3.97 | 8.11 | 73.99 | 26.57 | 70.57 | 25.76 | |
| RLOOBackbone Model=Qwen-2.5-VL-7B2025.10 | 21.73 | 6.58 | 5.12 | 0.0326 | 0.0705 | 5.1 | 10.56 | 79.67 | 24.79 | 75.8 | 24 |