Web Action Generation Efficiency on Mind2Web Cross-Website
364.1To_Pro Steps/TimeDuSAR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DuSARModel=Llama3.1-70B2025.12 | 364.1 | 11.8 | |
| DuSARModel=Gemma3-12B2025.12 | 408.3 | 18.5 | |
| DuSARModel=Llama3.1-8B2025.12 | 411 | 11.7 | |
| DuSARModel=Llama2-7B2025.12 | 551 | 79.3 | |
| ReActModel=Llama3.1-70B2025.12 | 1,477.3 | 315.6 | |
| ReActModel=Gemma3-12B2025.12 | 1,557.9 | 13.9 | |
| ReActModel=Llama3.1-8B2025.12 | 1,708.4 | 762.6 | |
| SynapseModel=Gemma3-12B2025.12 | 1,854.3 | 21.1 | |
| ReActModel=Llama2-7B2025.12 | 1,901 | 81.7 | |
| SynapseModel=Llama3.1-8B2025.12 | 2,059.8 | 56.6 | |
| SynapseModel=Llama3.1-70B2025.12 | 2,059.8 | 57.8 | |
| SynapseModel=Llama2-7B2025.12 | 2,078.9 | 41.3 | |
| TRADModel=Llama3.1-70B2025.12 | 3,284.8 | 31.1 | |
| TRADModel=Llama3.1-8B2025.12 | 3,323.3 | 548.2 | |
| TRADModel=Gemma3-12B2025.12 | 3,407.9 | 19.6 | |
| TRADModel=Llama2-7B2025.12 | 3,611.2 | 287.3 |