Path-level reasoning on UNOBench real (Hard)
79.5SR-PUNOGrasp
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| UNOGraspPrompting=Oracle (with SoM), Protocol=Standard2025.11 | 79.5 | 59.1 | 63.9 | 0.55 | |
| UNOGraspPrompting=Natural Language Prompting, Protocol=Standard2025.11 | 54.5 | 35.2 | 40.3 | 0.76 | |
| Gemini Robotics-ER 1.5Prompting=Natural Language Prompting, Protocol=Standard2025.11 | 38.9 | 33.3 | 35.2 | 0.89 | |
| Gemini Robotics-ER 1.5Prompting=Oracle (with SoM), Protocol=ICL2025.11 | 38.6 | 27.3 | 30.6 | 0.92 | |
| Qwen2.5-VLPrompting=Oracle (with SoM), Protocol=SFT2025.11 | 38.6 | 21.6 | 25.9 | — | |
| Gemini Robotics-ER 1.5Prompting=Natural Language Prompting, Protocol=ICL2025.11 | 38.6 | 27.3 | 30.6 | 0.92 | |
| Gemini Robotics-ER 1.5Prompting=Oracle (with SoM), Protocol=Standard2025.11 | 31.8 | 27.3 | 28.8 | 0.91 | |
| Qwen2.5-VLPrompting=Natural Language Prompting, Protocol=SFT2025.11 | 29.5 | 19.3 | 21.8 | — | |
| Qwen2.5-VLPrompting=Oracle (with SoM), Protocol=ICL2025.11 | 24.4 | 26.1 | 22.7 | 0.79 | |
| Qwen2.5-VLPrompting=Natural Language Prompting, Protocol=ICL2025.11 | 11.4 | 13.6 | 12.1 | 0.9 |