Robot Embodied Reasoning on Robot-R1 Bench
3.8ScoreXEmbodied
Evaluation Results
| Method | Links | |
|---|---|---|
| XEmbodiedvariant=Best2026.04 | 3.8 | |
| GPT-4o-2024-11-20Version=2024-11-202026.03 | 1.55 | |
| ROBOALIGN (Ours)Training=SFT + RL2026.03 | 1.38 | |
| Cosmos-Reason1-7BParameters=7B2026.03 | 1.19 | |
| RoboBrain2.0-7BVersion=2.0, Parameters=7B2026.03 | 1.17 | |
| ROBOALIGN w/o RLTraining=SFT only2026.03 | 1.14 | |
| Qwen2.5-VL-7B-InsParameters=7B, Variant=Instruct2026.03 | 1.02 | |
| VeBrain-8BParameters=8B2026.03 | 1.02 | |
| GPT-4o2026.04 | 0.35 | |
| Qwen2.5-VL-7B2026.04 | 0.25 | |
| Qwen3-VL-A3B-30B2026.04 | 0.05 | |
| Qwen3-VL-32B2026.04 | 0.05 | |
| Cosmos-R12026.04 | 0.05 | |
| Qwen2.5-VL-32B2026.04 | 0 | |
| Mimo-Embodied2026.04 | 0 |