Interactive Agent Tasks on Mind2Web
18.86Task Success RateMNL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MNLBase Model=DeepSeek-V3.2-Exp2025.12 | 18.86 | 67.55 | 12 | 395 | |
| ACEBase Model=DeepSeek-V3.2-Exp2025.12 | 15.82 | 57.8 | 580 | 58,602 | |
| Vanilla modelBase Model=DeepSeek-V3.2-Exp2025.12 | 15.49 | 66.32 | — | — | |
| MNLBase Model=Qwen3-8B2025.12 | 2.02 | 15.64 | 695 | 556 | |
| Vanilla modelBase Model=Qwen3-8B2025.12 | 1.35 | 11.54 | — | — | |
| MementoBase Model=DeepSeek-V3.2-Exp2025.12 | 0.34 | 12.6 | 1,707 | 4,822 | |
| MementoBase Model=Qwen3-8B2025.12 | 0 | 0.18 | 1,707 | 4,749 | |
| ACEBase Model=Qwen3-8B2025.12 | 0 | 0 | 363 | 24,284 |