Multi-objective Continuous Control on MuJoCo HalfCheetah (test)
66.95MERDense Oracle
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Dense OracleAccess=Privileged (dynamic event signals and hand-crafted rules)2026.03 | 66.95 | 99.98 | |
| DPI-PPOAccess=Standard environment state, Inference Mode=Infer preferences purely from interaction2026.03 | 42.1 | 81 | |
| ENVELOPE-PPOAccess=Standard environment state2026.03 | 13.52 | 53.28 |