Offline Action Consistency Diagnostic on AITW
74.2Action AccuracyQwen2.5-VL Agent + VisCritic
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-VL Agent + VisCriticbase_agent=Qwen2.5-VL Agent, critic_module=VisCritic2026.06 | 74.2 | |
| Qwen2.5-VL Agent + GUI-Critic-R1base_agent=Qwen2.5-VL Agent, critic_module=GUI-Critic-R12026.06 | 73.3 | |
| Qwen2.5-VL Agent + GuidNavbase_agent=Qwen2.5-VL Agent, critic_module=GuidNav2026.06 | 73 | |
| Qwen2.5-VL Agent + GUI-PRAbase_agent=Qwen2.5-VL Agent, critic_module=GUI-PRA2026.06 | 72.5 | |
| Qwen2.5-VL Agentbase_agent=Qwen2.5-VL Agent2026.06 | 70.4 | |
| ShowUI + VisCriticbase_agent=ShowUI, critic_module=VisCritic2026.06 | 69.8 | |
| ShowUI + GUI-PRAbase_agent=ShowUI, critic_module=GUI-PRA2026.06 | 68.8 | |
| CogAgent + VisCriticbase_agent=CogAgent, critic_module=VisCritic2026.06 | 67.1 | |
| ShowUIbase_agent=ShowUI2026.06 | 66.7 | |
| SeeClick + GUI-Critic-R1base_agent=SeeClick, critic_module=GUI-Critic-R12026.06 | 64.8 | |
| CogAgentbase_agent=CogAgent2026.06 | 64.8 | |
| SeeClick + GUI-PRAbase_agent=SeeClick, critic_module=GUI-PRA2026.06 | 64.5 | |
| SeeClick + VisCriticbase_agent=SeeClick, critic_module=VisCritic2026.06 | 64.5 | |
| SeeClick + BacktrackAgentbase_agent=SeeClick, critic_module=BacktrackAgent2026.06 | 64.2 | |
| SeeClickbase_agent=SeeClick2026.06 | 62.1 |