Manipulation Task on EmbodiedBench Vanilla
28.3Base ScoreSceneDiver
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SceneDiverbackbone=gemini-2.5-flash2026.06 | 28.3 | 17.9 | 26.3 | 28.3 | 33.3 | |
| gemini-2.5-flashvariant=base2026.06 | 20.8 | 14.6 | 22.9 | 24.2 | 23.3 | |
| SceneDiverbackbone=Qwen2.5-VL-7B-Instruct-AWQ2026.06 | 12.5 | 10.4 | 12.9 | 16.3 | 16.7 | |
| Qwen2.5-VL-7B-Instruct-AWQvariant=base2026.06 | 6.7 | 8.8 | 10.4 | 15 | 8.3 |