List your company

OS-level Agent Task Execution on OSWorld subset

46.48Success RateEVA-CUA-8B
22.549628.762334.97541.1877Jun 10, 2026

Evaluation results

MethodLinks
EVA-CUA-8BDeployment Mode=teacher-free test-time policy2026.0646.48——
SFT collector with SRCDeployment Mode=teacher-assisted data collection2026.0643.2716.825.29
Expert SFT + SRCDeployment Mode=teacher-free test-time policy2026.0640.1511.86—
Expert SFTDeployment Mode=teacher-free test-time policy2026.0627.2212.81—
Base modelDeployment Mode=teacher-free test-time policy2026.0623.4712.88—