Evaluator Accuracy on AndroidWorld
87.9Overall AccStepCritic
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| StepCriticModel=StepCritic2025.04 | 87.9 | 92.8 | 82.3 | |
| Captioner + GPT-4Model=Captioner + GPT-42025.04 | 84.6 | — | — | |
| Captioner + MixtralModel=Captioner + Mixtral2025.04 | 82.4 | — | — |