Web Agent Navigation on MiniWoB (full)
77.1Success RateGemini 3.1 Pro
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini 3.1 ProAgent harness=This work2026.04 | 77.1 | |
| Qwen3.5-27BParams=27B, Agent harness=This work2026.04 | 70.9 | |
| Claude 3.5 SonnetAgent harness=BrowserGym ecosystem2026.04 | 69.8 | |
| A3-Qwen3.5-9BParams=9B, Agent harness=This work2026.04 | 69 | |
| o1-miniAgent harness=BrowserGym ecosystem2026.04 | 67.8 | |
| Llama 3.1 405BParams=405B, Agent harness=BrowserGym ecosystem2026.04 | 64.6 | |
| GPT-4oAgent harness=BrowserGym ecosystem2026.04 | 63.8 | |
| Qwen3.5-9B (base)Params=9B, Agent harness=This work2026.04 | 63.2 | |
| Llama 3.1 70BParams=70B, Agent harness=BrowserGym ecosystem2026.04 | 57.6 | |
| GPT-4o-miniAgent harness=BrowserGym ecosystem2026.04 | 56.6 |