Policy Referral on GTAPolicy
9.45GTAPolicy ScoreTriMPI
Evaluation Results
| Method | Links | |
|---|---|---|
| TriMPIStrategy=PoRo-GRPO2025.10 | 9.45 | |
| TriMPIStrategy=PoRo-DAPO2025.10 | 9.13 | |
| CoT SFT + GRPORL Algorithm=GRPO2025.10 | 8.93 | |
| CoT SFT + DAPORL Algorithm=DAPO2025.10 | 8.73 | |
| CoT SFT2025.10 | 7.68 |