Web Agent Navigation on TimeWarp Overall Average v1:6
97.1Success RateHuman
Evaluation Results
| Method | Links | |
|---|---|---|
| Human2026.03 | 97.1 | |
| Qwen-3 4B (TimeWarp-BC)Protocol=TW, Type=Train2026.03 | 37.7 | |
| Qwen-3 4B Thinking (TimeWarp-BC)Protocol=TW, Type=Train2026.03 | 27.4 | |
| Llama-3.1 8B (TimeWarp-BC)Protocol=TW, Type=Train2026.03 | 27 | |
| Qwen-3VL 8BModality=HTML, Protocol=Zero-shot2026.03 | 24.3 | |
| Qwen-3 4BModality=HTML, Protocol=Zero-shot2026.03 | 20.5 | |
| Gemma-3 12BModality=HTML, Protocol=Zero-shot2026.03 | 15.3 | |
| Qwen-3VL 8B ThinkingModality=HTML, Protocol=Zero-shot2026.03 | 12.6 |