Web Navigation on Mind2Web Cross-domain
63.4Success Rate (Acc)PolySkill + Update
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PolySkill + UpdateBackbone=GPT-4.1, Approach Category=w/ Online Update2025.10 | 63.4 | 56 | |
| BaselineBackbone=GPT-4.12025.10 | 62.3 | — | |
| ASI + UpdateBackbone=GPT-4.1, Approach Category=w/ Online Update2025.10 | 62.1 | 66 | |
| PolySkillBackbone=GPT-4.1, Approach Category=Static Approaches2025.10 | 60.1 | 36 | |
| ASIBackbone=GPT-4.1, Approach Category=Static Approaches2025.10 | 57.3 | 33 | |
| GPT-4o-mini2025.06 | 9.33 | — | |
| Go-Browse-7BTraining Data=GO-BROWSE-WA (success trajectories), Base Model=Qwen-2.5-7B-Instruct2025.06 | 5.33 | — | |
| NNetNav-7BTraining Data=NNETNAV-WA2025.06 | 4 | — |