Multi-task reinforcement learning on Metaworld MT10 v2 (train/eval)
4.2TimeFAMO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FAMO2023.05 | 4.2 | — | |
| SDMGrad-OSS=12023.05 | 6.8 | — | |
| CAGrad-Fasttask_sampling_size=n=42023.05 | 8.6 | — | |
| Nash-MTLupdate_interval=every 100 iterations2023.05 | 9.3 | — | |
| Nash-MTLupdate_interval=every 50 iterations2023.05 | 9.7 | — | |
| SDMGrad-OSstrategy=objective sampling2023.05 | 9.7 | — | |
| MoCo2023.05 | 11.5 | — | |
| PCGrad2023.05 | 11.6 | — | |
| CAGrad2023.05 | 13.5 | — | |
| SDMGrad2023.05 | 13.6 | — | |
| Nash-MTLreport_source=reproduced2023.05 | 87.4 | — |