Embodied Question Answering and Decision Making on EQA-Decision Benchmark
81.55Static Scene AccuracyRoboDecision-8B
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| RoboDecision-8BModel Type=Embodied Baseline, Parameters=8B2026.05 | 81.55 | 70.82 | 68.12 | 52.95 | 65.02 | 69.93 | 68.06 | |
| Qwen3-VL-4B-ThinkingModel Type=General Baseline, Parameters=4B, Mode=Thinking2026.05 | 76.8 | 51.75 | 15.36 | 15.02 | 48.7 | 55.27 | 43.81 | |
| Qwen3-VL-8B-ThinkingModel Type=General Baseline, Parameters=8B, Mode=Thinking2026.05 | 76.65 | 55.42 | 23.14 | 16.72 | 51.75 | 55.75 | 46.57 | |
| Qwen3-VL-4B-InstructModel Type=General Baseline, Parameters=4B, Mode=Instruct2026.05 | 74.23 | 51.97 | 20.19 | 15.83 | 53.35 | 52.36 | 44.65 | |
| GPT-5Model Type=General Baseline2026.05 | 70.74 | 47.75 | 45.25 | 25.72 | 54.52 | 62.25 | 51.03 | |
| Qwen3-VL-8B-InstructModel Type=General Baseline, Parameters=8B, Mode=Instruct2026.05 | 70.47 | 54.84 | 35.51 | 23.98 | 54.02 | 54.27 | 48.84 | |
| Gemini-2.5-ProModel Type=General Baseline2026.05 | 66.7 | 56.54 | 47.56 | 17.42 | 56.35 | 47.56 | 48.68 | |
| RoboBrain-7B-2.0Model Type=Embodied Baseline, Parameters=7B, Version=2.02026.05 | 60.8 | 25.62 | 61.93 | 19.25 | 33.7 | 41.9 | 37.32 |