Visual Question Answering on VQA RoboSpatial ERQA EO Bench
46.7ERQAMiMo-Embodied (7B)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MiMo-Embodied (7B)Model Size=7B2026.06 | 46.7 | — | — | |
| + Detection qualityBase Model=Qwen3.5-4B, Training Data=Detection quality2026.06 | 45.6 | 67.4 | 37.9 | |
| SPARCBackbone=Qwen3.5-4B2026.06 | 44.9 | 62.4 | 35.3 | |
| + Base MixtureBase Model=Qwen3.5-4B, Training Data=Base Mixture2026.06 | 42.8 | 65.15 | 36.3 | |
| + EO-1.5MBase Model=Qwen3.5-4B, Training Data=EO-1.5M2026.06 | 42.5 | 69.45 | 42 | |
| Qwen3.5-4BModel Size=4B2026.06 | 30.3 | 40 | 30.5 | |
| RoboBrain 2.0 (7B)Model Size=7B2026.06 | 30.3 | — | — |