Web Agent Navigation on Mind2Web All 1.0
0.484Element AccuracyDuSAR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DuSARModel=Llama3.1-70B2025.12 | 0.484 | 0.381 | 400 | |
| DuSARModel=Gemma3-12B2025.12 | 0.461 | 0.381 | 372 | |
| DuSARModel=Llama3.1-8B2025.12 | 0.439 | 0.316 | 176 | |
| TRADModel=Llama3.1-70B2025.12 | 0.396 | 0.334 | 175 | |
| SynapseModel=Llama3.1-70B2025.12 | 0.313 | 0.194 | 54 | |
| TRADModel=Gemma3-12B2025.12 | 0.305 | 0.262 | 95 | |
| SynapseModel=Gemma3-12B2025.12 | 0.296 | 0.231 | 98 | |
| ReActModel=Gemma3-12B2025.12 | 0.218 | 0.156 | 43 | |
| SynapseModel=Llama3.1-8B2025.12 | 0.207 | 0.087 | 11 | |
| TRADModel=Llama3.1-8B2025.12 | 0.16 | 0.111 | 2 | |
| DuSARModel=Llama2-7B2025.12 | 0.135 | 0.097 | 4 | |
| SynapseModel=Llama2-7B2025.12 | 0.124 | 0.092 | 7 | |
| ReActModel=Llama3.1-70B2025.12 | 0.122 | 0.098 | 13 | |
| TRADModel=Llama2-7B2025.12 | 0.106 | 0.08 | 0 | |
| ReActModel=Llama2-7B2025.12 | 0.073 | 0.059 | 4 | |
| ReActModel=Llama3.1-8B2025.12 | 0.063 | 0.032 | 0 |