End-to-end task execution on OSWorld (test)
38.54Success RateR-WoM
Evaluation Results
| Method | Links | |
|---|---|---|
| R-WoMModel=Claude-3.7-Sonnet2025.10 | 38.54 | |
| R-WoMModel=Qwen-2.5-VL-72B2025.10 | 37.48 | |
| WebDreamerModel=Claude-3.7-Sonnet2025.10 | 31.24 | |
| RAGModel=Qwen-2.5-VL-72B2025.10 | 30.84 | |
| VanillaModel=Claude-3.7-Sonnet2025.10 | 28.47 | |
| WebDreamerModel=Qwen-2.5-VL-72B2025.10 | 28.37 | |
| RAGModel=Claude-3.7-Sonnet2025.10 | 27.76 | |
| VanillaModel=Qwen-2.5-VL-72B2025.10 | 26.36 | |
| R-WoMModel=Claude-3.5-Sonnet2025.10 | 26.01 | |
| WebDreamerModel=Claude-3.5-Sonnet2025.10 | 23.48 | |
| VanillaModel=Claude-3.5-Sonnet2025.10 | 22.43 | |
| RAGModel=Claude-3.5-Sonnet2025.10 | 22.19 |