Vision-Language-Action Learning on LIBERO Multi-task scenario
0.74FARCRL-VLA (V)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CRL-VLA (V)Goal-conditioned value=V-based2026.02 | 0.74 | 0.17 | 0 | |
| CRL-VLA (Q)Goal-conditioned value=Q-based2026.02 | 0.66 | -0.03 | 0 | |
| CRL-VLAvariant=V2026.02 | 0.63 | -0.1 | -0.7 | |
| LWFlearning_paradigm=Learning Without Forgetting2026.02 | 0.63 | 0.1 | 0 | |
| SLlearning_paradigm=Sequence Learning2026.02 | 0.62 | 0.07 | 0.25 | |
| ERlearning_paradigm=Experience Replay2026.02 | 0.62 | 0.05 | 0 | |
| SL2026.02 | 0.6 | 0 | -0.7 | |
| MTL2026.02 | 0.54 | 0.17 | -0.7 | |
| ER2026.02 | 0.52 | -0.05 | -0.65 | |
| MTLlearning_paradigm=Multi-Task Learning2026.02 | 0.49 | 0 | 0.25 | |
| LWF2026.02 | 0.43 | 0 | -0.65 | |
| CRL-VLAvariant=Q2026.02 | 0.33 | -0.45 | -0.7 |