Web Agent Navigation on Mind2Web OOD (test)
51.56Step AccuracySkillFlow
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SkillFlowMethod Type=Ours, Backbone Model=Qwen3.5-9B2026.05 | 51.56 | 84.89 | |
| FlowSteerMethod Type=Agent+RL, Backbone Model=Qwen3.5-9B2026.05 | 41.41 | 75.22 | |
| AgentFlowMethod Type=Agent+RL, Backbone Model=Qwen3.5-9B2026.05 | 39.84 | 71.77 | |
| SkillRLMethod Type=Agent+RL, Backbone Model=Qwen3.5-9B2026.05 | 35.16 | 69.54 | |
| AFlowMethod Type=Search-based workflow, Backbone Model=Qwen3.5-9B2026.05 | 30.47 | 62.29 | |
| v4-flashMethod Type=Baseline, Backbone Model=v4-flash2026.05 | 27.57 | 66.21 | |
| Qwen3.5Method Type=Baseline, Backbone Model=Qwen3.5-9B2026.05 | 26.64 | 63.85 | |
| Qwen3.5Method Type=GRPO, Backbone Model=Qwen3.5-9B2026.05 | 23.51 | 66.98 | |
| Qwen3.5Method Type=SFT, Backbone Model=Qwen3.5-9B2026.05 | 18.74 | 50.49 |