Success Rate Prediction on MMLU CF
100Perturbation Success Rate
TRUE (DAG-based predictor)
Updated 4d ago
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
2026.02 | 100 | 80 | 0.53 | - | 3.98 | |
2026.02 | 100 | 80 | - | 4.51 | - | |
2026.02 | 100 | 40 | 1.69 | - | 12.13 | |
2026.02 | 100 | 40 | - | 13.82 | - | |
2026.02 | 86.6 | 65 | 0.97 | - | 3.93 | |
2026.02 | 86.6 | 65 | - | 4.9 | - | |
2026.02 | 83.3 | 65 | 1 | - | -0.17 | |
2026.02 | 83.3 | 65 | - | 0.83 | - | |
2026.02 | 83.3 | 65 | 0.98 | - | -0.15 | |
2026.02 | 83.3 | 65 | - | 0.83 | - | |
2026.02 | 66.6 | 75 | 0.67 | - | 3.84 | |
2026.02 | 66.6 | 75 | - | 4.51 | - |