Deformable Linear Object Manipulation on DLO-Lab
100CoilingSAPO
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| SAPO2026.06 | 100 | 0 | 0 | 100 | 0 | 80 | 0 | 0 | 35 | |
| GD2026.06 | 100 | 0 | 0 | 100 | 0 | 0 | 0 | 0 | 25 | |
| CMA-ES2026.06 | 100 | 100 | 87 | 100 | 93 | 93 | 93 | 27 | 86.6 | |
| SAPOTraining Protocol=per-task single-task policies, Evaluation Trajectories=15 per task2026.06 | 100 | — | — | 100 | — | 80 | 0 | — | 70 | |
| SHAC2026.06 | 93 | 0 | 0 | 100 | 0 | 73 | 0 | 0 | 33.3 | |
| SHACTraining Protocol=per-task single-task policies, Evaluation Trajectories=15 per task2026.06 | 93 | — | — | 100 | — | 73 | 0 | — | 66.5 | |
| PPO2026.06 | 67 | 0 | 0 | 100 | 0 | 0 | 67 | 0 | 29.3 | |
| PPOTraining Protocol=per-task single-task policies, Evaluation Trajectories=15 per task2026.06 | 67 | — | — | 100 | — | 0 | 67 | — | 58.5 | |
| SmolVLA (multi-task)Training Protocol=multi-task (single policy trained jointly on all four tasks), Evaluation Trajectories=15 per task2026.06 | 60 | — | — | 100 | — | 73 | 7 | — | 60 | |
| SAC2026.06 | 0 | 0 | 0 | 100 | 33 | 0 | 0 | 0 | 16.6 | |
| SACTraining Protocol=per-task single-task policies, Evaluation Trajectories=15 per task2026.06 | 0 | — | — | 100 | — | 0 | 0 | — | 25 |