Reward Prediction on ManiSkill (image-based) 4-task Color Swapped (OOD)
96Pick ScoreReCouPLe-EC
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ReCouPLe-ECTask Setting=Multi-Task with Reasons, Lambda ratio=0.42026.03 | 96 | 100 | 100 | 98 | |
| ReCouPLe-ICTask Setting=Multi-Task with Reasons, Lambda ratio=0.42026.03 | 96 | 100 | 99.3 | 98.7 | |
| RFPTask Setting=Multi-Task with Reasons2026.03 | 80.7 | 96.7 | 94.7 | 83.3 | |
| BT-MultiTask Setting=Multi-Task2026.03 | 53.3 | 86.7 | 83.3 | 58.7 | |
| BTTask Setting=Single Task2026.03 | 16.7 | 61 | 46 | 5.3 |