| Environment | Robust RL (entropy) | Shaping-aware (entropy) | Δentropy | | Robust RL (KL) | Shaping-aware (KL) | ΔKL | |
|---|
| Hopper-v4 | 2905.12±133.86 | 3214.68±115.60 | +309.56 | ↑10.7% | 2882.08±186.12 | 3249.23±131.87 | +367.15 | ↑12.7% |
| Walker2d-v4 | 2861.81±316.32 | 3385.88±281.79 | +524.07 | ↑18.3% | 2676.94±263.54 | 3577.35±352.55 | +900.41 | ↑33.6% |
| HalfCheetah-v4 | 1386.73±21.84 | 1625.94±20.58 | +239.21 | ↑17.2% | 1583.30±35.38 | 3288.30±60.48 | +1705.00 | ↑107.7% |
| Humanoid-v4 | 463.15±25.82 | 505.54±15.78 | +42.39 | ↑9.2% | 372.60±28.52 | 418.10±13.90 | +45.50 | ↑12.2% |
| Reacher-v4 | −5.98±0.27 | −4.88±0.49 | +1.10 | ↑18.4% | −5.88±0.44 | −5.17±0.32 | +0.71 | ↑12.1% |
| Seaquest | 604.00±68.59 | 832.00±48.99 | +228.00 | ↑37.7% | 616.00±39.19 | 840.00±29.39 | +224.00 | ↑36.4% |