Embodied Manipulation on Guava Simulation OOD Long Horizon
93.3Success Rate: Separate Food and UtensilsGuava-Agent-4B
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Guava-Agent-4BEpisodes=152026.06 | 93.3 | — | 93.3 | 6.7 | 0 | |
| CaP-Agent0Episodes=152026.06 | 86.7 | — | 93.3 | 0 | 53.3 | |
| GPT-5.4Episodes=152026.06 | 86.7 | — | 93.3 | 46.7 | 80 | |
| Qwen3.5-4BEpisodes=152026.06 | 0 | — | 20 | 53.3 | 0 |