RoboRewardBench
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Reward Modeling | RoboRewardBench | MAE0.665 | 24 | |
| Reward Prediction | RoboRewardBench 1.0 | Overall MAE0.665 | 12 | |
| Reward Prediction | RoboRewardBench | MAE0.67 | 8 | |
| Robotic Manipulation Trajectory Preference Prediction | RoboRewardBench 500 randomly sampled trajectory pairs | Accuracy87.4 | 5 | |
| Value-Order Correlation | RoboRewardBench (held-out) | Spearman VOC0.966 | 4 |
Showing 5 of 5 rows