Web Navigation Success on VisualWebArena
19.8Success RateGPT-4o
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4oInput Modality=with text2025.05 | 19.8 | |
| ViGoRL-7bBackbone=7b2025.05 | 11.2 | |
| ICAL-7BBackbone=7B, Input Modality=with text2025.05 | 8.2 | |
| ViGoRL-3bBackbone=3b2025.05 | 6.4 | |
| Qwen2.5-VL-7BBackbone=7B2025.05 | 5.5 | |
| Qwen2.5-VL-3B + Vanilla GRPOBackbone=3B, Training Protocol=Vanilla GRPO2025.05 | 4.8 | |
| Qwen2.5-VL-3B + SFT directBackbone=3B, Training Protocol=SFT direct2025.05 | 4.5 | |
| Qwen2.5-VL-3BBackbone=3B2025.05 | 4.2 | |
| Qwen2-VL-7BBackbone=7B, Input Modality=with text2025.05 | 2.9 |