AD Trigger threshold control on Monte Carlo (MC) (held-out 20%)
0.3MAEGFPO-F
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GFPO-F2026.06 | 0.3 | 0.7 | 100 | 95.442 | 27.926 | 95.442 | 27.926 | |
| L-GRPO2026.06 | 1 | 2.2 | 100 | 95.369 | 27.764 | 95.369 | 27.764 | |
| DQN2026.06 | 1.1 | 2.3 | 99.3 | 95.352 | 27.717 | 95.357 | 27.739 | |
| PPO2026.06 | 1.1 | 2.2 | 100 | 95.353 | 27.726 | 95.353 | 27.726 | |
| ADT2026.06 | 1.1 | 2.2 | 100 | 95.353 | 27.726 | 95.353 | 27.726 | |
| CPO2026.06 | 1.1 | 2.2 | 100 | 95.353 | 27.726 | 95.353 | 27.726 | |
| GRPO2026.06 | 1.3 | 2.5 | 95.8 | 95.347 | 27.733 | 95.356 | 27.74 | |
| DQN-F2026.06 | 1.6 | 3 | 84.7 | 95.543 | 28.185 | 95.474 | 27.948 | |
| GFPO-FR2026.06 | 1.9 | 2.4 | 97.9 | 95.842 | 28.875 | 95.842 | 28.868 | |
| PID2026.06 | 2 | 5.1 | 72.9 | 95.233 | 27.298 | 94.99 | 26.287 | |
| DSPOT2026.06 | 2.2 | 6.1 | 66.7 | 94.992 | 26.767 | 94.867 | 26.05 | |
| Constant2026.06 | 4.2 | 13.1 | 47.9 | 94.279 | 25.404 | 94.758 | 25.285 |