Loco-navigation Goal-conditioned Reinforcement Learning on humanoidmaze expert
92Success RateSAW + TTT
Evaluation Results
| Method | Links | |
|---|---|---|
| SAW + TTTBackbone RL Algorithm=SAW, Test-time Training (TTT) variant=GC-TTT2025.07 | 92 | |
| SAW + TTT (no critic)Backbone RL Algorithm=SAW, Test-time Training (TTT) variant=GC-TTT (no critic)2025.07 | 87 | |
| SAWBackbone RL Algorithm=SAW, Test-time Training (TTT) variant=None (Pre-trained)2025.07 | 85 | |
| GC-IQL-DDPG + TTTBackbone RL Algorithm=GC-IQL-DDPG, Test-time Training (TTT) variant=GC-TTT2025.07 | 56 | |
| GC-IQL-DDPG + TTT (no critic)Backbone RL Algorithm=GC-IQL-DDPG, Test-time Training (TTT) variant=GC-TTT (no critic)2025.07 | 40 | |
| GC-IQL-DDPGBackbone RL Algorithm=GC-IQL-DDPG, Test-time Training (TTT) variant=None (Pre-trained)2025.07 | 32 | |
| GC-BC + TTTBackbone RL Algorithm=GC-BC, Test-time Training (TTT) variant=GC-TTT2025.07 | 21 | |
| GC-BC + TTT (no critic)Backbone RL Algorithm=GC-BC, Test-time Training (TTT) variant=GC-TTT (no critic)2025.07 | 15 | |
| GC-BCBackbone RL Algorithm=GC-BC, Test-time Training (TTT) variant=None (Pre-trained)2025.07 | 7 |