Robotic Manipulation on ManiSkill3
96.4Average Success RateIPI
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| IPIBase Model=OpenVLA-7B, Framework=StARe2025.12 | 96.4 | 94.3 | 97.3 | 98.5 | 95.5 | — | |
| StA-PPOBase Model=OpenVLA-7B, Framework=StARe2025.12 | 93.4 | 92.7 | 96 | 95.3 | 89.7 | — | |
| IPIBase Model=Pi0.5, Framework=StARe2025.12 | 89.9 | 84.3 | 99.3 | 95 | 80.7 | — | |
| StA-PPOBase Model=Pi0.5, Framework=StARe2025.12 | 86.7 | 80.7 | 98 | 92.7 | 75.3 | — | |
| RoboFAC-7B2025.12 | 82.7 | 85.5 | 80.4 | 80.7 | 84 | — | |
| pi_RLBase Model=Pi0.52025.12 | 80.1 | 72.3 | 96.7 | 93.3 | 58 | — | |
| RL4VLABase Model=OpenVLA-7B2025.12 | 70.5 | 64 | 95.7 | 90.3 | 32 | — | |
| Env RewardReward Model=Env Reward2026.03 | 66.87 | — | — | — | — | — | |
| Structured 4D Latent Predictive Model2026.07 | 61.3 | 84 | — | 84 | 16 | — | |
| Task Completion RewardReward Model=Task Completion Reward2026.03 | 60.93 | — | — | — | — | — | |
| Temporal Contrastive RewardReward Model=Temporal Contrastive Reward2026.03 | 60.31 | — | — | — | — | — | |
| Absolute Progress RewardReward Model=Absolute Progress Reward2026.03 | 60 | — | — | — | — | — | |
| RoboReward-8BReward Model=RoboReward-8B2026.03 | 59.06 | — | — | — | — | — | |
| Robometer-4BReward Model=Robometer-4B2026.03 | 56.56 | — | — | — | — | — | |
| DP2026.07 | 55.7 | 56 | — | 87 | 24 | — | |
| SmolVLAProtocol=fine-tuning2025.12 | 51.5 | 12.7 | 86.3 | 90.7 | 16.3 | — | |
| DP32026.07 | 49.3 | 47 | — | 94 | 7 | — | |
| OctoProtocol=fine-tuning2025.12 | 39.5 | 1 | 67 | 90 | 0 | — | |
| UniPi2026.07 | 29.3 | 35 | — | 49 | 4 | — | |
| StA-TPOBase Model=Pi0.5, Trajectory samples=100, Trajectory preference pairs=50, Framework=StARe2025.12 | 27.7 | 28 | 22.3 | 44.7 | 15.7 | — | |
| SFTBase Model=Pi0.5, Trajectory samples=1002025.12 | 25.6 | 26.3 | 18.3 | 43 | 10.7 | — | |
| TesserAct2026.07 | 24.7 | 31 | — | 43 | 3 | — | |
| DiT w/ RoboFlow4Dmodule=RoboFlow4D2026.05 | 23.7 | 12 | 45 | — | — | 14 | |
| GRAPEBase Model=Pi0.52025.12 | 22.6 | 22.7 | 16.3 | 45 | 6.3 | — | |
| DP w/ RoboFlow4Dmodule=RoboFlow4D2026.05 | 22 | 12 | 41 | — | — | 13 | |
| StA-TPOBase Model=OpenVLA-7B, Trajectory samples=100, Trajectory preference pairs=50, Framework=StARe2025.12 | 20.8 | 19.3 | 16 | 35.7 | 12.3 | — | |
| GRAPEBase Model=OpenVLA-7B2025.12 | 18 | 15.7 | 13.3 | 35.3 | 7.7 | — | |
| SFTBase Model=OpenVLA-7B, Trajectory samples=1002025.12 | 15 | 12 | 11.7 | 31 | 5.3 | — | |
| DiT2026.05 | 12.7 | 2 | 32 | — | — | 4 | |
| DP2026.05 | 12.3 | 3 | 31 | — | — | 3 | |
| ACT2026.05 | 11 | 1 | 30 | — | — | 2 | |
| BC2026.05 | 9 | 0 | 26 | — | — | 1 | |
| OpenVLA2026.05 | 4 | 2 | 10 | — | — | 0 |