Robot Manipulation Aggregate on Franka Manipulation (Real-World Evaluation)
84Mean Success RateAFRO
Evaluation results
| Method | Links | |
|---|---|---|
| AFROPre-training Domain=Out-of-domain, Evaluation Protocol=Pre-train & FT, Backbone=PointTransformer [60]2025.11 | 84 | |
| SafeDojoType=Ours2026.06 | 76 | |
| AFROPre-training Domain=In-domain, Evaluation Protocol=Standard, Backbone=PointTransformer [60]2025.11 | 70 | |
| WoVRType=Model-based RL2026.06 | 66 | |
| PointDifPre-training Domain=Out-of-domain, Evaluation Protocol=Pre-train & FT, Backbone=PointTransformer [60]2025.11 | 65 | |
| DynaMo-3DPre-training Domain=Out-of-domain, Evaluation Protocol=Pre-train & FT, Backbone=PointTransformer [60]2025.11 | 58 | |
| PointDifPre-training Domain=In-domain, Evaluation Protocol=Standard, Backbone=PointTransformer [60]2025.11 | 53 | |
| WMPOType=Model-based RL2026.06 | 52 | |
| DP3Pre-training Domain=In-domain, Evaluation Protocol=Standard, Backbone=PointTransformer [60]2025.11 | 50 | |
| DynaMo-3DPre-training Domain=In-domain, Evaluation Protocol=Standard, Backbone=PointTransformer [60]2025.11 | 48 | |
| FVPPre-training Domain=Out-of-domain, Evaluation Protocol=Pre-train & FT, Backbone=PointTransformer [60]2025.11 | 46 | |
| SafeVLAType=Model-free RL2026.06 | 40 | |
| PPOType=Model-free RL2026.06 | 36 | |
| FVPPre-training Domain=In-domain, Evaluation Protocol=Standard, Backbone=PointTransformer [60]2025.11 | 34 | |
| OpenVLA-OFTType=SFT2026.06 | 20 | |
| AEGISType=CBF2026.06 | 18 |