GUI Agent Planning and Execution on WebArena
70Success Rate (Gitlab)WEBXSKILL (Grounded)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| WEBXSKILL (Grounded)Model=GPT-52026.04 | 70 | 57.7 | 100 | 69.5 | 65.9 | 65.7 | |
| WEBXSKILL + SkillWeaverModel=GPT-52026.04 | 66.7 | 53.9 | 68.4 | 59.7 | 56.8 | 57.1 | |
| VanillaModel=GPT-52026.04 | 63.3 | 50 | 79 | 59.7 | 50 | 65.7 | |
| WEBXSKILL (Guided)Model=GPT-52026.04 | 63.3 | 61.5 | 89.5 | 68.8 | 59.1 | 80 | |
| Vanilla + MAPModel=GPT-52026.04 | 60 | 46.2 | 84.2 | 63 | 52.3 | 80 | |
| WEBXSKILL + WALTModel=GPT-52026.04 | 56.7 | 50 | 84.2 | 62.3 | 52.3 | 77.1 | |
| WEBXSKILL + SkillWeaverModel=Qwen2026.04 | 53.3 | 30.8 | 63.2 | 43.5 | 36.4 | 42.9 | |
| WALTModel=GPT-52026.04 | 51.8 | 28.1 | 57.7 | 42.9 | 30.9 | 47.4 | |
| WALTModel=Qwen2026.04 | 50 | 46.2 | 57.9 | 44.2 | 31.8 | 45.7 | |
| WEBXSKILL + WALTModel=Qwen2026.04 | 50 | 46.2 | 57.9 | 44.2 | 31.8 | 45.7 | |
| WEBXSKILL (Guided)Model=Qwen2026.04 | 50 | 57.7 | 63.2 | 53.9 | 47.7 | 57.1 | |
| WEBXSKILL (Grounded)Model=Qwen2026.04 | 46.7 | 38.5 | 47.4 | 48.7 | 47.7 | 60 | |
| SkillWeaverModel=GPT-52026.04 | 44.4 | 41.3 | 30.2 | 37.4 | 28.3 | 42.9 | |
| VanillaModel=Qwen2026.04 | 43.3 | 42.3 | 57.9 | 45.5 | 40.9 | 48.6 | |
| Vanilla + MAPModel=Qwen2026.04 | 43.3 | 46.2 | 73.7 | 47.1 | 38.6 | 47.1 | |
| SkillWeaverModel=Qwen2026.04 | 40 | 46.2 | 66.7 | 43.8 | 37.8 | 38.7 | |
| Zero-ShotBase Model=GPT-4o [20]2026.03 | 32.14 | 6.42 | 27.36 | 22.82 | — | — | |
| HATSBase Model=InternVL2-8B [4]2026.03 | 27.81 | 24.77 | 20.19 | 24.87 | — | — | |
| HATSBase Model=Qwen2-VL-7B [61]2026.03 | 25 | 21.1 | 22.77 | 23.28 | — | — | |
| HATSBase Model=InternVL2-4B [4]2026.03 | 22.73 | 20.19 | 17.92 | 20.6 | — | — | |
| OS-GenesisBase Model=Qwen2-VL-7B [61]2026.03 | 15.87 | 5 | 15.63 | 12.81 | — | — | |
| Task-DrivenBase Model=InternVL2-4B [4]2026.03 | 9.52 | 5 | 0 | 5.72 | — | — | |
| Task-Driven w. Self-InstructBase Model=InternVL2-4B [4]2026.03 | 9.52 | 7.5 | 0 | 6.41 | — | — | |
| OS-GenesisBase Model=InternVL2-4B [4]2026.03 | 7.94 | 7.5 | 3.13 | 6.53 | — | — | |
| Task-Driven w. Self-InstructBase Model=InternVL2-8B [4]2026.03 | 7.94 | 0 | 6.25 | 5.3 | — | — | |
| Task-DrivenBase Model=InternVL2-8B [4]2026.03 | 6.35 | 2.5 | 0 | 3.58 | — | — | |
| OS-GenesisBase Model=InternVL2-8B [4]2026.03 | 6.35 | 10 | 9.34 | 8.16 | — | — | |
| Task-DrivenBase Model=Qwen2-VL-7B [61]2026.03 | 6.35 | 5 | 6.25 | 5.95 | — | — | |
| Task-Driven w. Self-InstructBase Model=Qwen2-VL-7B [61]2026.03 | 4.84 | 7.5 | 3.13 | 5.12 | — | — | |
| Zero-ShotBase Model=InternVL2-8B [4]2026.03 | 1.37 | 0.95 | 0.97 | 1.13 | — | — | |
| Zero-ShotBase Model=Qwen2-VL-7B [61]2026.03 | 1.14 | 0.98 | 1.96 | 1.32 | — | — | |
| Zero-ShotBase Model=InternVL2-4B [4]2026.03 | 1.11 | 0 | 0.94 | 0.76 | — | — |