List your company
ResearchFormerly WizWand

Object-level reasoning on UNOBench Hard synthetic (test)

67.6OPGemini Robotics-ER 1.5
1.14418.39735.6552.903Nov 28, 2025
Updated 8mo ago

Evaluation results

MethodLinks
Gemini Robotics-ER 1.5Prompting Strategy=Oracle (with SoM), Evaluation Mode=In-Context Learning (ICL)2025.1167.624.333
UNOGraspPrompting Strategy=Oracle (with SoM)2025.116343.648.7
Gemini Robotics-ER 1.5Prompting Strategy=Oracle (with SoM), Evaluation Mode=Base Prompting2025.1146.114.420.8
UNOGraspPrompting Strategy=Natural Language Prompting2025.1144.620.125.7
Qwen2.5-VLPrompting Strategy=Oracle (with SoM), Evaluation Mode=In-Context Learning (ICL)2025.1113.910.810.6
Gemini Robotics-ER 1.5Prompting Strategy=Natural Language Prompting, Evaluation Mode=Base Prompting2025.117.823.1
Gemini Robotics-ER 1.5Prompting Strategy=Natural Language Prompting, Evaluation Mode=In-Context Learning (ICL)2025.116.11.72.6
Qwen2.5-VLPrompting Strategy=Natural Language Prompting, Evaluation Mode=In-Context Learning (ICL)2025.113.72.22.7