List your company
ResearchFormerly WizWand

Reward Prediction on UI-TARS-7B trajectories

96.5AccuracyPRORE
83.70887.02990.3593.671Sep 26, 2025
Updated 6mo ago

Evaluation results

MethodLinks
PRORELast N States=Proactive2025.0996.566.7
WebRLLast N States=Full2025.0993.922.2
WebRLLast N States=12025.099320
Step-CriticLast N States=12025.099320
Step-CriticLast N States=Full2025.099320
DistRLLast N States=Full2025.0989.514.3
DigiRLLast N States=Full2025.0986.811.7
DistRLLast N States=12025.0984.225
DigiRLLast N States=12025.0984.210