Embodied Continual Learning on LIBERO Long 1.0 (test)
78.6FWTT-Stellar
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| T-Stellarsource=reported by Stellar VLA2026.06 | 78.6 | 44.3 | 41.2 | 31.9 | |
| TS-Stellarsource=reported by Stellar VLA2026.06 | 74.7 | 37.9 | 41.7 | 36.8 | |
| ERbackbone=π0, replay=true2026.06 | 74.4 | 22.8 | 56.2 | 51.8 | |
| Task-MoEbackbone=π0, architecture=modular experts2026.06 | 72.8 | 12 | 63.8 | 57.8 | |
| UniVLAsource=reported by Stellar VLA2026.06 | 70.6 | 32.4 | 45 | 17 | |
| SCEbackbone=π02026.06 | 69.6 | 0.5 | 70 | 73.4 | |
| MoDEsource=reported by Stellar VLA2026.06 | 68.3 | 45.7 | 30.6 | 18.7 | |
| Sequentialbackbone=π0, replay=none2026.06 | 62.4 | 58.4 | 14.8 | 4 |