Reinforcement Learning on Reach online downstream setting
1Normalized RewardOracle
Evaluation results
| Method | Links | ||
|---|---|---|---|
| OracleEvaluation Protocol=Online Downstream2026.05 | 1 | 0.033 | |
| Task-OnlyEvaluation Protocol=Online Downstream2026.05 | 0.96 | 1 | |
| Safe-VPLEvaluation Protocol=Online Downstream2026.05 | 0.91 | 0.124 | |
| Safe-CPLEvaluation Protocol=Online Downstream2026.05 | 0.9 | 0.074 | |
| SOPLEvaluation Protocol=Online Downstream2026.05 | 0.76 | 0.083 | |
| RCEvaluation Protocol=Online Downstream, omega=0.52026.05 | 0.73 | 0.081 |