Embodied AI Task Success on EmbodiedBench Navigation (test)
62.5Success Rate (Base Task)SVA
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SVABackbone=Qwen2.5-9B2026.07 | 62.5 | 59.72 | 52.78 | 47.22 | 58.33 | 56.11 | |
| SVABackbone=Qwen2.5-27B2026.07 | 61.11 | 56.94 | 58.33 | 58.33 | 52.78 | 57.5 | |
| SVABackbone=Qwen2.5-4B2026.07 | 54.17 | 41.67 | 45.83 | 41.67 | 50 | 46.67 | |
| SVABackbone=GPT-4o2026.07 | 51.39 | 50 | 45.83 | 41.67 | 50 | 47.78 | |
| BaseBackbone=GPT-4o2026.07 | 47.22 | 30.56 | 51.39 | 38.89 | 44.44 | 42.5 | |
| BaseBackbone=Qwen2.5-4B2026.07 | 45.83 | 38.89 | 31.94 | 31.94 | 25 | 34.72 | |
| SVABackbone=Gemma-2-9B-it2026.07 | 41.67 | 37.5 | 50 | 38.89 | 37.5 | 41.11 | |
| BaseBackbone=Qwen2.5-9B2026.07 | 40.28 | 37.5 | 40.28 | 30.56 | 47.22 | 39.17 | |
| BaseBackbone=Qwen2.5-27B2026.07 | 40.28 | 51.39 | 44.44 | 37.5 | 16.67 | 38.06 | |
| BaseBackbone=Gemma-2-9B-it2026.07 | 37.5 | 25 | 38.89 | 30.56 | 11.11 | 28.61 |